google.yml

# yaml-language-server: $schema=.schema.json
name: Google
id: google
pricing_urls:
  - https://ai.google.dev/gemini-api/docs/pricing
  - https://cloud.google.com/vertex-ai/generative-ai/pricing
  # https://cloud.google.com/vertex-ai/generative-ai/pricing#context-caching explains how caching is calculated
  # Grounding with Google Search (`web_searches_kcount`): Gemini 3.x bills each unique non-empty search query at $14/1k,
  # Gemini 2.x bills each grounded prompt at $35/1k however many queries it ran, and only when it returned a web source.
  # Callers report the billed count as `web_searches`. Free monthly/daily allowances are account-wide and not modelled.
  # https://ai.google.dev/gemini-api/docs/google-search#pricing
api_pattern: 'https://(.*\.)?googleapis\.com'
model_match:
  contains: gemini
provider_match:
  or:
    - contains: google
    - contains: vertex
    - contains: gemini

# https://github.com/googleapis/nodejs-vertexai/blob/v1.10.0/src/types/content.ts#L465
# A previous comment said:
# > Note, this suggests `promptTokenCount` and `candidatesTokenCount` are optional, but they seem to be always present
# Then we found a case where `candidatesTokenCount` was missing, so we assume `promptTokenCount` can be missing too.
extractors:
  - root: usageMetadata
    model_path: modelVersion
    mappings:
      - path: promptTokenCount
        dest: input_tokens
        required: false
      - path: cachedContentTokenCount
        dest: cache_read_tokens
        required: false
      - path:
          - cacheTokensDetails
          - type: array-match
            field: modality
            match:
              equals: TEXT
          - tokenCount
        dest: cache_text_read_tokens
        required: false
      - path:
          - cacheTokensDetails
          - type: array-match
            field: modality
            match:
              equals: AUDIO
          - tokenCount
        dest: cache_audio_read_tokens
        required: false
      - path:
          - cacheTokensDetails
          - type: array-match
            field: modality
            match:
              equals: IMAGE
          - tokenCount
        dest: cache_image_read_tokens
        required: false
      - path:
          - cacheTokensDetails
          - type: array-match
            field: modality
            match:
              equals: DOCUMENT
          - tokenCount
        dest: cache_image_read_tokens
        required: false
      - path:
          - cacheTokensDetails
          - type: array-match
            field: modality
            match:
              equals: VIDEO
          - tokenCount
        dest: cache_video_read_tokens
        required: false
      - path:
          - promptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: TEXT
          - tokenCount
        dest: input_text_tokens
        required: false
      - path:
          - promptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: AUDIO
          - tokenCount
        dest: input_audio_tokens
        required: false
      - path:
          - promptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: IMAGE
          - tokenCount
        dest: input_image_tokens
        required: false
      - path:
          - promptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: DOCUMENT
          - tokenCount
        dest: input_image_tokens
        required: false
      - path:
          - promptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: VIDEO
          - tokenCount
        dest: input_video_tokens
        required: false
      - path:
          - candidatesTokensDetails
          - type: array-match
            field: modality
            match:
              equals: TEXT
          - tokenCount
        dest: output_text_tokens
        required: false
      - path:
          - candidatesTokensDetails
          - type: array-match
            field: modality
            match:
              equals: AUDIO
          - tokenCount
        dest: output_audio_tokens
        required: false
      - path:
          - candidatesTokensDetails
          - type: array-match
            field: modality
            match:
              equals: IMAGE
          - tokenCount
        dest: output_image_tokens
        required: false
      - path:
          - candidatesTokensDetails
          - type: array-match
            field: modality
            match:
              equals: DOCUMENT
          - tokenCount
        dest: output_image_tokens
        required: false
      - path:
          - candidatesTokensDetails
          - type: array-match
            field: modality
            match:
              equals: VIDEO
          - tokenCount
        dest: output_video_tokens
        required: false
      - path: candidatesTokenCount
        dest: output_tokens
        required: false
      - path: thoughtsTokenCount
        dest: output_tokens
        required: false
      - path: thoughtsTokenCount
        dest: output_reasoning_tokens
        required: false
      - path: toolUsePromptTokenCount
        dest: input_tokens
        required: false
      - path: toolUsePromptTokenCount
        dest: input_tool_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: TEXT
          - tokenCount
        dest: input_text_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: TEXT
          - tokenCount
        dest: input_text_tool_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: AUDIO
          - tokenCount
        dest: input_audio_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: AUDIO
          - tokenCount
        dest: input_audio_tool_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: IMAGE
          - tokenCount
        dest: input_image_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: IMAGE
          - tokenCount
        dest: input_image_tool_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: DOCUMENT
          - tokenCount
        dest: input_image_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: DOCUMENT
          - tokenCount
        dest: input_image_tool_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: VIDEO
          - tokenCount
        dest: input_video_tokens
        required: false
      - path:
          - toolUsePromptTokensDetails
          - type: array-match
            field: modality
            match:
              equals: VIDEO
          - tokenCount
        dest: input_video_tool_tokens
        required: false
  # https://github.com/anthropics/anthropic-sdk-python/blob/v0.62.0/src/anthropic/types/usage.py#L12
  - root: usage
    api_flavor: anthropic
    mappings:
      - path: input_tokens
        dest: input_tokens
      - path: cache_creation_input_tokens
        dest: input_tokens
        required: false
      - path: cache_read_input_tokens
        dest: input_tokens
        required: false
      - path: cache_creation_input_tokens
        dest: cache_write_tokens
        required: false
      - path: [cache_creation, ephemeral_5m_input_tokens]
        dest: cache_write_5m_tokens
        required: false
      - path: [cache_creation, ephemeral_1h_input_tokens]
        dest: cache_write_1h_tokens
        required: false
      - path: cache_read_input_tokens
        dest: cache_read_tokens
        required: false
      - path: output_tokens
        dest: output_tokens
  - api_flavor: chat # openai
    root: usage
    mappings:
      - path: prompt_tokens
        dest: input_tokens
      - path: [prompt_tokens_details, cached_tokens]
        dest: cache_read_tokens
        required: false
      - path: [prompt_tokens_details, audio_tokens]
        dest: input_audio_tokens
        required: false
      - path: [completion_tokens_details, audio_tokens]
        dest: output_audio_tokens
        required: false
      - path: [completion_tokens_details, reasoning_tokens]
        dest: output_reasoning_tokens
        required: false
      - path: completion_tokens
        dest: output_tokens

fallback_model_providers:
  - anthropic

models:
  - id: claude-3-5-haiku
    canonical_model: anthropic/claude-3-5-haiku-latest
    match:
      contains: claude-3-5-haiku
    context_window: 200000
    prices_checked: 2025-07-05
    prices:
      input_mtok: 0.8
      cache_write_mtok: 1
      cache_read_mtok: 0.08
      output_mtok: 4

  - id: claude-3-5-sonnet
    canonical_model: anthropic/claude-3-5-sonnet
    # this model also covers claude-3-5-sonnet-v2
    match:
      contains: claude-3-5-sonnet
    context_window: 200000
    prices_checked: 2025-07-05
    prices:
      input_mtok: 3
      cache_write_mtok: 3.75
      cache_read_mtok: 0.3
      output_mtok: 15

  - id: claude-3-7-sonnet
    canonical_model: anthropic/claude-3-7-sonnet-latest
    match:
      contains: claude-3-7-sonnet
    context_window: 200000
    prices_checked: 2025-07-05
    prices:
      input_mtok: 3
      cache_write_mtok: 3.75
      cache_read_mtok: 0.3
      output_mtok: 15

  - id: claude-3-haiku
    canonical_model: anthropic/claude-3-haiku
    match:
      contains: claude-3-haiku
    context_window: 200000
    prices_checked: 2025-07-05
    prices:
      input_mtok: 0.25
      cache_write_mtok: 0.3
      cache_read_mtok: 0.03
      output_mtok: 1.25

  - id: claude-3-opus
    canonical_model: anthropic/claude-3-opus-latest
    match:
      contains: claude-3-opus
    prices_checked: 2025-07-05
    prices:
      input_mtok: 15
      cache_write_mtok: 18.75
      cache_read_mtok: 1.5
      output_mtok: 75

  - id: claude-4-opus
    match:
      or:
        - contains: claude-4-opus
        - contains: claude-opus-4@
        - contains: claude-opus-4-0
        - contains: claude-opus-4-1
        - equals: claude-opus-4
    context_window: 200000
    prices_checked: 2025-07-05
    prices:
      input_mtok: 15
      cache_write_mtok: 18.75
      cache_read_mtok: 1.5
      output_mtok: 75

  - id: claude-4-sonnet
    canonical_model: anthropic/claude-sonnet-4-0
    match:
      or:
        - contains: claude-4-sonnet
        - contains: claude-sonnet-4-0
        - contains: claude-sonnet-4@
        - contains: claude-sonnet-4-2025
        - equals: claude-sonnet-4
        - equals: anthropic/claude-sonnet-4
    context_window: 200000
    prices_checked: 2025-07-05
    prices:
      input_mtok: 3
      cache_write_mtok: 3.75
      cache_read_mtok: 0.3
      output_mtok: 15

  - id: claude-fable-5
    canonical_model: anthropic/claude-fable-5
    match:
      or:
        - ends_with: claude-fable-5
        - contains: claude-fable-5@
    context_window: 1000000
    prices_checked: 2026-09-25
    price_comments: >-
      Flat pricing across full 1M context window.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models
      The one-hour cache write rate is the global endpoint's, from the same pricing page.
    prices:
      input_mtok: 10
      cache_read_mtok: 1.0
      cache_write_mtok: 12.5
      cache_write_1h_mtok: 20
      output_mtok: 50

  - id: claude-fable-5-1
    canonical_model: anthropic/claude-fable-5-1
    match:
      or:
        - ends_with: claude-fable-5-1
        - contains: claude-fable-5-1@
    context_window: 1000000
    prices_checked: 2026-09-25
    price_comments: >-
      Flat pricing across full 1M context window. Global endpoint; regional and
      multi-region endpoints add a 10% premium.
      Cache hits are 0.025x base input (not the usual 0.1x), unique to Fable 5.1 and Mythos 5.1.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models
      Model ID ref: https://platform.claude.com/docs/en/build-with-claude/claude-on-vertex-ai
      The one-hour cache write rate is the global endpoint's, from the same pricing page.
    prices:
      input_mtok: 10
      cache_read_mtok: 0.25
      cache_write_mtok: 12.5
      cache_write_1h_mtok: 20
      output_mtok: 50

  - id: claude-haiku-4-5
    canonical_model: anthropic/claude-haiku-4-5
    match:
      or:
        - contains: claude-haiku-4-5
        - contains: claude-haiku-4.5
    context_window: 200000
    prices_checked: 2026-10-15
    price_comments: >-
      Vertex AI Global endpoint rates from https://cloud.google.com/vertex-ai/generative-ai/pricing
      (Claude models, Global tab): input $1 / 1M tokens, output $5 / 1M tokens, 5-minute cache writes
      $1.25 / 1M tokens, 1-hour cache writes $2 / 1M tokens, cache hits $0.10 / 1M tokens. The Vertex
      table also publishes batch rates (half of input and output); batch variants are not modelled for
      Vertex Claude entries in this file.
    prices:
      input_mtok: 1
      cache_read_mtok: 0.1
      cache_write_mtok: 1.25
      cache_write_1h_mtok: 2
      output_mtok: 5

  - id: claude-haiku-5-5
    canonical_model: anthropic/claude-haiku-5-5
    match:
      or:
        - contains: claude-haiku-5-5
        - contains: claude-haiku-5.5
        - contains: claude-5-5-haiku
        - contains: claude-5.5-haiku
    context_window: 1000000
    prices_checked: 2026-10-08
    price_comments: >-
      Global endpoint pricing; multi-region and regional endpoints carry a 10% premium.
      Prompts over 100K input tokens bill every token at the long-context rate.
      Rates match Anthropic's list price and OpenRouter's Google endpoint.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models
      OpenRouter ref: https://openrouter.ai/api/v1/models/anthropic/claude-haiku-5.5/endpoints
    prices:
      input_mtok:
        base: 0.1
        tiers:
          - start: 100000
            price: 0.5
      cache_write_mtok:
        base: 0.125
        tiers:
          - start: 100000
            price: 0.625
      cache_write_1h_mtok:
        base: 0.2
        tiers:
          - start: 100000
            price: 1
      cache_read_mtok:
        base: 0.01
        tiers:
          - start: 100000
            price: 0.05
      output_mtok:
        base: 0.5
        tiers:
          - start: 100000
            price: 2.5

  - id: claude-opus-4-5
    canonical_model: anthropic/claude-opus-4-5
    match:
      or:
        - contains: claude-4-5-opus
        - contains: claude-opus-4-5
        - contains: claude-4.5-opus
        - contains: claude-opus-4.5
    context_window: 200000
    prices_checked: 2026-10-12
    price_comments: >-
      A 200K context window, so there is no long-context tier.
      Global endpoint rates; regional and multi-region endpoints carry a 10% premium.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models
      Model ID ref: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/use-claude
    prices:
      input_mtok: 5
      cache_read_mtok: 0.50
      cache_write_mtok: 6.25
      cache_write_1h_mtok: 10
      output_mtok: 25

  - id: claude-opus-4-6
    canonical_model: anthropic/claude-opus-4-6
    # Google documents 1M for this exact model: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/opus-4-6
    match:
      or:
        - contains: claude-4-6-opus
        - contains: claude-opus-4-6
        - contains: claude-4.6-opus
        - contains: claude-opus-4.6
    prices_checked: 2026-09-25
    price_comments: >-
      Flat pricing across the full 1M context window since the long-context premium was removed on
      2026-03-13, including on Vertex AI: https://claude.com/blog/1m-context-ga
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models (global endpoint)
    prices:
      - prices:
          input_mtok:
            base: 5
            tiers:
              - start: 200000
                price: 10
          cache_read_mtok:
            base: 0.50
            tiers:
              - start: 200000
                price: 1
          cache_write_mtok:
            base: 6.25
            tiers:
              - start: 200000
                price: 12.50
          output_mtok:
            base: 25
            tiers:
              - start: 200000
                price: 37.50
      - constraint:
          start_date: 2026-03-13
        prices:
          input_mtok: 5
          cache_read_mtok: 0.50
          cache_write_mtok: 6.25
          cache_write_1h_mtok: 10
          output_mtok: 25

  - id: claude-opus-4-7
    canonical_model: anthropic/claude-opus-4-7
    match:
      or:
        - contains: claude-4-7-opus
        - contains: claude-opus-4-7
        - contains: claude-4.7-opus
        - contains: claude-opus-4.7
    context_window: 1000000
    prices_checked: 2026-09-25
    price_comments: >-
      Flat pricing across full 1M context window.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models
      The one-hour cache write rate is the global endpoint's, from the same pricing page.
    prices:
      input_mtok: 5
      cache_read_mtok: 0.50
      cache_write_mtok: 6.25
      cache_write_1h_mtok: 10
      output_mtok: 25

  - id: claude-opus-4-8
    canonical_model: anthropic/claude-opus-4-8
    match:
      or:
        - contains: claude-4-8-opus
        - contains: claude-opus-4-8
        - contains: claude-4.8-opus
        - contains: claude-opus-4.8
    context_window: 1000000
    prices_checked: 2026-09-25
    price_comments: >-
      Flat pricing across full 1M context window.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models
      The one-hour cache write rate is the global endpoint's, from the same pricing page.
    prices:
      input_mtok: 5
      cache_read_mtok: 0.50
      cache_write_mtok: 6.25
      cache_write_1h_mtok: 10
      output_mtok: 25

  - id: claude-opus-5
    canonical_model: anthropic/claude-opus-5
    match:
      or:
        - contains: claude-5-opus
        - ends_with: claude-opus-5
        - contains: claude-opus-5@
        - contains: claude-5.0-opus
        - contains: claude-opus-5.0
    context_window: 1000000
    prices_checked: 2026-09-25
    price_comments: >-
      Global endpoint pricing, flat across the full 1M context window. Multi-region and regional
      endpoints carry a 10% premium.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models
      The one-hour cache write rate is the global endpoint's, from the same pricing page.
    prices:
      input_mtok: 5
      cache_read_mtok: 0.50
      cache_write_mtok: 6.25
      cache_write_1h_mtok: 10
      output_mtok: 25

  - id: claude-opus-5-5
    canonical_model: anthropic/claude-opus-5-5
    match:
      or:
        - contains: claude-5-5-opus
        - ends_with: claude-opus-5-5
        - contains: claude-opus-5-5@
        - contains: claude-5.5-opus
        - contains: claude-opus-5.5
    context_window: 1000000
    prices_checked: 2026-09-25
    price_comments: >-
      Global endpoint pricing, flat across the full 1M context window. Multi-region and regional
      endpoints carry a 10% premium. Google's pricing page renders client-side and could not be read
      when this entry was added; the rates follow Anthropic's list price, as every other Claude entry
      in this file does. Cache hits are 0.05x base input (not the usual 0.1x), as on Sonnet 5.5.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models
      Anthropic ref: https://platform.claude.com/docs/en/about-claude/pricing#model-pricing
      Model ID ref: https://platform.claude.com/docs/en/models/opus-5-5/overview
      The one-hour cache write rate is the global endpoint's, from the same pricing page.
    prices:
      input_mtok: 4
      cache_read_mtok: 0.20
      cache_write_mtok: 5
      cache_write_1h_mtok: 8
      output_mtok: 20

  - id: claude-sonnet-4-5
    canonical_model: anthropic/claude-sonnet-4-5
    match:
      or:
        - contains: claude-sonnet-4-5
        - contains: claude-sonnet-4.5
    # Google documents 200K as the GA context and 1M as beta:
    # https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/sonnet-4-5
    context_window: 200000
    prices_checked: 2026-09-25
    price_comments: >-
      Long-context rates apply above 200K input tokens.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#partner-models
      The one-hour cache write rate is the global endpoint's, from the same pricing page.
    prices:
      input_mtok:
        base: 3
        tiers:
          - start: 200000
            price: 6
      cache_write_mtok:
        base: 3.75
        tiers:
          - start: 200000
            price: 7.5
      cache_write_1h_mtok:
        base: 6
        tiers:
          - start: 200000
            price: 12
      cache_read_mtok:
        base: 0.3
        tiers:
          - start: 200000
            price: 0.6
      output_mtok:
        base: 15
        tiers:
          - start: 200000
            price: 22.5

  - id: claude-sonnet-4-6
    canonical_model: anthropic/claude-sonnet-4-6
    match:
      or:
        - contains: claude-sonnet-4-6
        - contains: claude-sonnet-4.6
    # Google documents a 1M context window:
    # https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/claude/sonnet-4-6
    prices_checked: 2026-09-25
    price_comments: >-
      Flat pricing across the full 1M context window.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#partner-models
      The one-hour cache write rate is the global endpoint's, from the same pricing page.
    prices:
      input_mtok: 3
      cache_write_mtok: 3.75
      cache_write_1h_mtok: 6
      cache_read_mtok: 0.3
      output_mtok: 15

  - id: claude-sonnet-5-5
    canonical_model: anthropic/claude-sonnet-5-5
    match:
      or:
        - contains: claude-sonnet-5-5
        - contains: claude-sonnet-5.5
        - contains: claude-5-5-sonnet
        - contains: claude-5.5-sonnet
    context_window: 1000000
    prices_checked: 2026-10-08
    price_comments: >-
      Global endpoint pricing, flat across the full 1M context window. Multi-region and regional
      endpoints carry a 10% premium. Rates match Anthropic's list price and OpenRouter's Google endpoint.
      Cache hits are 0.05x base input (not the usual 0.1x), as on Opus 5.5.
      Ref: https://cloud.google.com/vertex-ai/generative-ai/pricing#claude-models
      Anthropic ref: https://platform.claude.com/docs/en/about-claude/pricing#model-pricing
      OpenRouter ref: https://openrouter.ai/api/v1/models/anthropic/claude-sonnet-5.5/endpoints
    prices:
      input_mtok: 2
      cache_write_mtok: 2.5
      cache_write_1h_mtok: 4
      cache_read_mtok: 0.1
      output_mtok: 10

  - id: gemini-1.0-pro-vision-001
    name: gemini 1.0 pro vision
    description: >-
      Google's first-generation advanced multimodal model that can understand
      text, code, and images. It provides strong reasoning capabilities and
      follows instructions effectively.
    match:
      equals: gemini-1.0-pro-vision-001
    context_window: 32768
    price_comments: I can't find anything about this model or it's pricing, so trusting the original source
    prices_checked: 2025-07-04
    prices:
      input_mtok: 0.125
      output_mtok: 0.375

  - id: gemini-1.5-flash
    name: gemini 1.5 flash
    description: >-
      A faster, more cost-effective variant of Gemini 1.5 that maintains strong
      capabilities while optimizing for performance and cost efficiency.
      Suitable for production deployments requiring high throughput.
    match:
      contains: gemini-1.5-flash
    context_window: 1000000
    prices_checked: 2025-07-04
    prices:
      input_mtok:
        base: 0.075
        tiers:
          - start: 128000
            price: 0.15
      output_mtok:
        base: 0.3
        tiers:
          - start: 128000
            price: 0.6
      cache_read_mtok:
        base: 0.01875
        tiers:
          - start: 128000
            price: 0.0375

  - id: gemini-1.5-pro
    name: gemini 1.5 Pro
    description: >-
      Google's most capable multimodal model with an extremely long context
      window of up to 1 million tokens. It excels at complex reasoning,
      long-form content processing, and multimodal understanding.
    match:
      contains: gemini-1.5-pro
    context_window: 1000000
    prices_checked: 2025-06-22
    prices:
      input_mtok:
        base: 1.25
        tiers:
          - start: 128000
            price: 2.50
      output_mtok:
        base: 5.00
        tiers:
          - start: 128000
            price: 10.00

  - id: gemini-2.0-flash
    name: gemini 2.0 flash
    description: >-
      The newest generation of Google's Gemini models, featuring improved
      reasoning, instruction following, and factual accuracy, with the Flash
      variant optimized for cost-efficiency and performance.
    match:
      or:
        - ends_with: gemini-2.0-flash
        - contains: gemini-2.0-flash-0
        - contains: gemini-2.0-flash-exp
        - contains: gemini-2.0-flash-thinking
        - contains: gemini-2.0-flash-latest
    context_window: 1000000
    deprecated: true
    prices_checked: 2026-10-15
    price_comments: >-
      Listed as "(Shut down)" under "Previous models" in the Gemini API models catalog
      (https://ai.google.dev/gemini-api/docs/models, catalog last updated 2026-10-13). Checked 2026-10-15;
      kept for historical usage.
    prices:
      input_mtok: 0.1
      input_audio_mtok: 0.7
      output_mtok: 0.4
      cache_read_mtok: 0.025
      cache_audio_read_mtok: 0.175
      web_searches_kcount: 35 # per grounded prompt, not per query

  - id: gemini-2.0-flash-lite
    name: gemini 2.0 flash lite
    description: >-
      A lighter, more cost-effective version of Gemini 2.0 Flash, designed for
      applications requiring high efficiency while maintaining good performance.
      Ideal for high-volume, cost-sensitive deployments.
    match:
      contains: gemini-2.0-flash-lite
    context_window: 1000000
    deprecated: true
    prices_checked: 2026-10-15
    price_comments: >-
      Listed as "(Shut down)" under "Previous models" in the Gemini API models catalog
      (https://ai.google.dev/gemini-api/docs/models, catalog last updated 2026-10-13). Checked 2026-10-15;
      kept for historical usage.
    prices:
      input_mtok: 0.075
      output_mtok: 0.3

  - id: gemini-2.5-flash
    name: Gemini 2.5 Flash
    description: >-
      Gemini 2.5 Flash is Google's state-of-the-art workhorse model, specifically designed for advanced reasoning, coding,
      mathematics, and scientific tasks. It includes built-in "thinking" capabilities,
      enabling it to provide responses with greater accuracy and nuanced context handling.
    match:
      or:
        - equals: gemini-2.5-flash
        - equals: gemini-2.5-flash-latest
        - equals: gemini-2.5-flash-preview-09-2025
    context_window: 1048576
    prices_checked: 2026-10-14
    price_comments: >-
      Batch rates verified on 2026-10-14 at https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history
      before then is not modelled. Batch halves input and output; context caching stays at the standard rate.
    prices:
      input_mtok: 0.3
      input_audio_mtok: 1
      cache_read_mtok: 0.03
      cache_audio_read_mtok: 0.1
      output_mtok: 2.5
      web_searches_kcount: 35 # per grounded prompt, not per query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.15
          input_audio_mtok: 0.5
          output_mtok: 1.25

  - id: gemini-2.5-flash-image
    name: Gemini 2.5 Flash Image
    description: >-
      Google's specialized image generation model optimized for fast, high-quality image generation.
      Outputs images at 1024x1024 resolution, with each image consuming 1290 output tokens.
    match:
      or:
        - equals: gemini-2.5-flash-image
        - equals: gemini-2.5-flash-image-preview
    context_window: 1000000
    deprecated: true
    prices_checked: 2026-10-10
    price_comments: >-
      Deprecated. Gemini API pricing page announces "Gemini 2.5 Flash Image / Nano Banana (DEPRECATED — shut
      down Oct 2, 2026); migrate to Gemini 3.1 Flash Image or 3.1 Flash Lite Image."
      (https://ai.google.dev/gemini-api/docs/pricing). Image output is priced at $30 per 1M tokens, with each
      1024x1024 image = 1290 tokens = $0.039/image. Cache pricing is not available for this
      model. Text output uses the Gemini 2.5 Flash $2.50 rate and is the aggregate remainder when modality details omit text.
    prices:
      input_mtok: 0.3
      output_mtok: 2.5
      output_image_mtok: 30

  - id: gemini-2.5-flash-lite
    name: Gemini 2.5 Flash Lite
    description: >-
      Gemini 2.5 Flash-Lite is a lightweight reasoning model in the Gemini 2.5 family, optimized for ultra-low latency and
      cost efficiency. It offers improved throughput, faster token generation, and better
      performance across common benchmarks compared to earlier Flash models. By default, "thinking" (i.e. multi-pass reasoning)
      is disabled to prioritize speed, but developers can enable it via the Reasoning
      API parameter to selectively trade off cost for intelligence.
    match:
      or:
        - equals: gemini-2.5-flash-lite
        - and:
            - starts_with: gemini-2.5-flash-lite-preview
            - regex: "^(?!.*-[tT][tT][sS](?:$|-))"
    context_window: 1000000
    prices_checked: 2026-10-14
    price_comments: >-
      Batch rates verified on 2026-10-14 at https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history
      before then is not modelled. Batch halves input and output; context caching stays at the standard rate.
    prices:
      input_mtok: 0.1
      input_audio_mtok: 0.3
      cache_read_mtok: 0.01
      cache_audio_read_mtok: 0.03
      output_mtok: 0.4
      web_searches_kcount: 35 # per grounded prompt, not per query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.05
          input_audio_mtok: 0.15
          output_mtok: 0.2

  - id: gemini-2.5-flash-lite-preview-tts
    name: Gemini 2.5 Flash-Lite Preview TTS
    match:
      equals: gemini-2.5-flash-lite-preview-tts
    context_window: 8192
    prices_checked: 2026-08-26
    price_comments: See https://cloud.google.com/text-to-speech/pricing#gemini-tts.
    prices:
      input_mtok: 0.5
      output_mtok: 10

  - id: gemini-2.5-flash-preview
    name: Gemini 2.5 Flash Preview 05-20
    description: >-
      Gemini 2.5 Flash May 20th Checkpoint is Google's state-of-the-art workhorse model, specifically designed for advanced
      reasoning, coding, mathematics, and scientific tasks. It includes built-in "thinking"
      capabilities, enabling it to provide responses with greater accuracy and nuanced context handling.
    match:
      or:
        - contains: gemini-2.5-flash-preview-05-20
        - contains: gemini-2.5-flash-preview-04-17
        - equals: gemini-2.5-flash-preview-05-20:thinking
        - equals: gemini-2.5-flash-preview
        - equals: gemini-2.5-flash-preview:thinking
    price_comments: from https://cloud.google.com/vertex-ai/generative-ai/pricing should be retired 2025-07-15
    deprecated: true
    prices_checked: 2026-09-29
    prices:
      input_mtok: 0.15
      output_mtok: 0.6
      web_searches_kcount: 35 # per grounded prompt, not per query

  - id: gemini-2.5-flash-tts
    name: Gemini 2.5 Flash TTS
    match:
      or:
        - equals: gemini-2.5-flash-tts
        - equals: gemini-2.5-flash-preview-tts
    context_window: 8192
    prices_checked: 2026-10-14
    price_comments: >-
      See https://cloud.google.com/text-to-speech/pricing#gemini-tts. Batch rates verified on 2026-10-14 at
      https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is not modelled. Rates are
      from the Gemini API gemini-2.5-flash-preview-tts Batch tab.
    prices:
      input_mtok: 0.5
      output_mtok: 10
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.25
          output_mtok: 5

  - id: gemini-2.5-pro
    name: Gemini 2.5 Pro
    description: >-
      Gemini 2.5 Pro is Google's state-of-the-art AI model designed for advanced reasoning, coding, mathematics, and scientific
      tasks. It employs "thinking" capabilities, enabling it to reason through responses
      with enhanced accuracy and nuanced context handling. Gemini 2.5 Pro achieves top-tier performance on multiple benchmarks,
      including first-place positioning on the LMArena leaderboard, reflecting superior
      human-preference alignment and complex problem-solving abilities.
    match:
      and:
        - starts_with: gemini-2.5-pro
        - regex: "^(?!.*-[tT][tT][sS](?:$|-))"
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-pro. Batch rates verified on 2026-10-14 at
      https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is not modelled. Batch halves
      input and output in both prompt-size tiers; context caching stays at the standard rate.
    prices_checked: 2026-10-14
    prices:
      input_mtok:
        base: 1.25
        tiers:
          - start: 200000
            price: 2.5
      cache_read_mtok:
        base: 0.125
        tiers:
          - start: 200000
            price: 0.25
      output_mtok:
        base: 10
        tiers:
          - start: 200000
            price: 15
      web_searches_kcount: 35 # per grounded prompt, not per query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok:
            base: 0.625
            tiers:
              - start: 200000
                price: 1.25
          output_mtok:
            base: 5
            tiers:
              - start: 200000
                price: 7.5

  - id: gemini-2.5-pro-tts
    name: Gemini 2.5 Pro TTS
    match:
      or:
        - equals: gemini-2.5-pro-tts
        - equals: gemini-2.5-pro-preview-tts
    context_window: 8192
    prices_checked: 2026-10-14
    price_comments: >-
      See https://cloud.google.com/text-to-speech/pricing#gemini-tts. Batch rates verified on 2026-10-14 at
      https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is not modelled. Rates are
      from the Gemini API gemini-2.5-pro-preview-tts Batch tab.
    prices:
      input_mtok: 1
      output_mtok: 20
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.5
          output_mtok: 10

  - id: gemini-3-flash-preview
    name: Gemini 3 Flash Preview
    description: >-
      Google's ultra-fast frontier model optimized for speed and efficiency. Delivers state-of-the-art
      performance while maintaining low latency and cost, with improved reasoning and coding capabilities.
    match:
      or:
        - equals: gemini-3-flash-preview
        - starts_with: gemini-3-flash-preview-
    context_window: 1000000
    prices_checked: 2026-10-14
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Standard pricing shown; Batch API offers 50% discount on
      input/output. Batch rates verified on 2026-10-14 at https://ai.google.dev/gemini-api/docs/pricing (Batch tab);
      batch history before then is not modelled. Batch halves input and output; context caching is "Same as Standard".
    prices:
      - prices:
          input_mtok: 0.5
          input_audio_mtok: 1
          cache_read_mtok: 0.05
          cache_audio_read_mtok: 0.1
          output_mtok: 3
      - constraint:
          # Vertex AI pricing: Gemini 3 grounding "Billing will start January 5, 2026"
          start_date: 2026-01-05
        prices:
          input_mtok: 0.5
          input_audio_mtok: 1
          cache_read_mtok: 0.05
          cache_audio_read_mtok: 0.1
          output_mtok: 3
          web_searches_kcount: 14 # per search query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.25
          input_audio_mtok: 0.5
          output_mtok: 1.5

  - id: gemini-3-pro-image-preview
    name: Gemini 3 Pro Image Preview
    description: >-
      Google's image generation model optimized for high-quality image generation.
      Supports 1K/2K and 4K resolution outputs with flexible pricing based on image dimensions.
    match:
      or:
        - equals: gemini-3-pro-image
        - starts_with: gemini-3-pro-image-preview
    context_window: 1000000
    prices_checked: 2026-10-14
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing#gemini-3-pro-image. Image output is priced at $120 per 1M
      tokens, with each 1K/2K image = 1120 tokens = $0.134/image and each 4K image = 2000 tokens = $0.24/image. Text and
      thinking output is $12 per 1M tokens and is the aggregate remainder when modality details omit text. Batch rates
      verified on 2026-10-14 at https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is
      not modelled. Batch input is $1 (text) and $0.0006 per 560-token image (~$1/1M); batch image output is listed per
      image only ($0.067 per 1K/2K image of 1120 tokens, $0.12 per 4K image of 2000 tokens), which is $60 per 1M tokens.
    prices:
      - prices:
          input_mtok: 2
          output_mtok: 12
          output_image_mtok: 120
      - constraint:
          # Vertex AI pricing: Gemini 3 grounding "Billing will start January 5, 2026"
          start_date: 2026-01-05
        prices:
          input_mtok: 2
          output_mtok: 12
          output_image_mtok: 120
          web_searches_kcount: 14 # per search query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 1
          output_mtok: 6
          output_image_mtok: 60

  - id: gemini-3-pro-preview
    name: Gemini 3 Pro Preview
    description:
      The best model in the world for multimodal understanding, and our most powerful agentic and vibe-coding
      model yet.
    match:
      or:
        - starts_with: gemini-3-pro-preview
        - equals: gemini-3-pro-text-preview
    deprecated: true
    prices_checked: 2026-10-15
    price_comments: >-
      Listed as "(Shut down)" under "Previous models" in the Gemini API models catalog
      (https://ai.google.dev/gemini-api/docs/models, catalog last updated 2026-10-13). Checked 2026-10-15;
      kept for historical usage.
    prices:
      - prices:
          input_mtok:
            base: 2
            tiers:
              - start: 200000
                price: 4
          cache_read_mtok:
            base: 0.2
            tiers:
              - start: 200000
                price: 0.4
          output_mtok:
            base: 12
            tiers:
              - start: 200000
                price: 18
      - constraint:
          # Vertex AI pricing: Gemini 3 grounding "Billing will start January 5, 2026"
          start_date: 2026-01-05
        prices:
          input_mtok:
            base: 2
            tiers:
              - start: 200000
                price: 4
          cache_read_mtok:
            base: 0.2
            tiers:
              - start: 200000
                price: 0.4
          output_mtok:
            base: 12
            tiers:
              - start: 200000
                price: 18
          web_searches_kcount: 14 # per search query

  - id: gemini-3.1-flash-image-preview
    name: Gemini 3.1 Flash Image Preview
    description: >-
      Google's latest image generation model (Nano Banana 2) optimized for fast, high-quality image generation.
      Supports multiple output resolutions from 512px to 4K, with text and thinking output priced separately
      from image output tokens.
    match:
      or:
        - equals: gemini-3.1-flash-image
        - starts_with: gemini-3.1-flash-image-preview
    context_window: 1000000
    prices_checked: 2026-10-14
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Text and thinking output is priced at $3 per 1M tokens and is
      the aggregate remainder when modality details omit text. Image output is priced at $60 per 1M tokens. Preview
      model - pricing may change. Batch rates verified on 2026-10-14 at https://ai.google.dev/gemini-api/docs/pricing
      (Batch tab); batch history before then is not modelled. Batch halves input, text output and image output.
    prices:
      input_mtok: 0.5
      output_mtok: 3
      output_image_mtok: 60
      web_searches_kcount: 14 # per search query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.25
          output_mtok: 1.5
          output_image_mtok: 30

  - id: gemini-3.1-flash-lite
    name: Gemini 3.1 Flash Lite
    description: >-
      Google's fastest and most cost-efficient Gemini 3 series model, built for intelligence at scale.
      Optimized for high-volume, low-latency applications while maintaining strong multimodal capabilities.
    match:
      or:
        - equals: gemini-3.1-flash-lite
        - starts_with: gemini-3.1-flash-lite-preview
    context_window: 1000000
    prices_checked: 2026-10-14
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Batch rates verified on 2026-10-14 at
      https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is not modelled. Batch halves
      input, output and context caching.
    prices:
      input_mtok: 0.25
      input_audio_mtok: 0.5
      cache_read_mtok: 0.025
      cache_audio_read_mtok: 0.05
      output_mtok: 1.5
      web_searches_kcount: 14 # per search query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.125
          input_audio_mtok: 0.25
          cache_read_mtok: 0.0125
          cache_audio_read_mtok: 0.025
          output_mtok: 0.75

  - id: gemini-3.1-flash-lite-image
    name: Gemini 3.1 Flash Lite Image
    description: >-
      Google's low-latency, cost-efficient image generation and editing model.
    match:
      starts_with: gemini-3.1-flash-lite-image
    prices_checked: 2026-10-14
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-lite-image. Batch rates verified on 2026-10-14
      at https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is not modelled. Batch
      halves input, text output and image output.
    prices:
      input_mtok: 0.25
      output_mtok: 1.5
      output_image_mtok: 30
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.125
          output_mtok: 0.75
          output_image_mtok: 15

  - id: gemini-3.1-flash-live-preview
    name: Gemini 3.1 Flash Live Preview
    description: Google's low-latency audio-to-audio model with multimodal input support.
    match:
      starts_with: gemini-3.1-flash-live-preview
    prices_checked: 2026-09-29
    price_comments: See https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-live-preview.
    prices:
      input_mtok: 0.75
      input_audio_mtok: 3
      input_image_mtok: 1
      input_video_mtok: 1
      output_mtok: 4.5
      output_audio_mtok: 12
      web_searches_kcount: 14 # per search query

  - id: gemini-3.1-flash-tts-preview
    name: Gemini 3.1 Flash TTS Preview
    description: Preview token-priced text-to-speech model in the Gemini 3.1 Flash family.
    match:
      equals: gemini-3.1-flash-tts-preview
    prices_checked: 2026-10-15
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Token-priced TTS rates; batch is half input and output.
    prices:
      input_mtok: 1
      output_mtok: 20
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-15 }
        prices:
          input_mtok: 0.5
          output_mtok: 10

  - id: gemini-3.1-pro-preview
    name: Gemini 3.1 Pro Preview
    description: The latest performance, intelligence, and usability improvements to the best model family in the world
      for multimodal understanding, agentic capabilities, and vibe-coding.
    match:
      or:
        - starts_with: gemini-3.1-pro-preview
        # Family alias: the Gemini API's `modelVersion` resolved it to this model on 2026-09-29.
        # Move it when Google points it at a newer model.
        - equals: gemini-pro-latest
    prices_checked: 2026-10-14
    price_comments: >-
      Batch rates verified on 2026-10-14 at https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history
      before then is not modelled. Batch halves input and output in both prompt-size tiers; context caching is "Same as
      Standard".
    prices:
      input_mtok:
        base: 2
        tiers:
          - start: 200000
            price: 4
      cache_read_mtok:
        base: 0.2
        tiers:
          - start: 200000
            price: 0.4
      output_mtok:
        base: 12
        tiers:
          - start: 200000
            price: 18
      web_searches_kcount: 14 # per search query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok:
            base: 1
            tiers:
              - start: 200000
                price: 2
          output_mtok:
            base: 6
            tiers:
              - start: 200000
                price: 9

  - id: gemini-3.5-flash
    name: Gemini 3.5 Flash
    description: >-
      Google's most intelligent model built for speed, combining frontier intelligence with
      improved reasoning, coding, and multimodal understanding.
    match:
      or:
        - equals: gemini-3.5-flash
        - starts_with: gemini-3.5-flash-preview
        - regex: '^gemini-3\.5-flash-\d'
    context_window: 1000000
    prices_checked: 2026-09-29
    price_comments:
      See https://ai.google.dev/gemini-api/docs/pricing. Standard tier pricing shown; Batch and Flex tiers
      offer 50% discount on input/output.
    prices:
      input_mtok: 1.5
      cache_read_mtok: 0.15
      output_mtok: 9
      web_searches_kcount: 14 # per search query

  - id: gemini-3.5-flash-lite
    name: Gemini 3.5 Flash Lite
    description: >-
      Google's fastest and most cost-efficient Gemini 3.5 series model, optimized for high-volume,
      low-latency applications while maintaining strong multimodal capabilities.
    match:
      or:
        - starts_with: gemini-3.5-flash-lite
        # Family alias: the Gemini API's `modelVersion` resolved it to this model on 2026-09-29.
        # Move it when Google points it at a newer model.
        - equals: gemini-flash-lite-latest
    context_window: 1000000
    prices_checked: 2026-10-14
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Standard tier pricing shown; Batch and Flex tiers offer 50%
      discount. Input rate is unified across text/image/video/audio (no separate audio rate). Batch rates verified on
      2026-10-14 at https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is not
      modelled. Batch halves input and output; batch context caching is $0.02, not half of the $0.03 standard rate.
    prices:
      input_mtok: 0.3
      cache_read_mtok: 0.03
      output_mtok: 2.5
      web_searches_kcount: 14 # per search query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.15
          cache_read_mtok: 0.02
          output_mtok: 1.25

  - id: gemini-3.6-flash
    name: Gemini 3.6 Flash
    description: >-
      Google's most intelligent model built for speed, combining frontier intelligence with improved
      reasoning, coding, and multimodal understanding.
    match:
      starts_with: gemini-3.6-flash
    context_window: 1000000
    prices_checked: 2026-10-14
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Standard tier pricing shown; Batch and Flex tiers offer 50%
      discount. No separate audio input rate documented. Introductory rates apply through 2026-12-31. The cache storage
      price ($0.50 per 1M tokens per hour) has no unit in prices/units.yml and is not represented here. Batch rates
      verified on 2026-10-14 at https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is
      not modelled. Batch halves input, output and context caching, at both the introductory and the 2027-01-01 rates.
    prices:
      - prices:
          input_mtok: 0.75
          cache_read_mtok: 0.075
          output_mtok: 3.75
          web_searches_kcount: 14 # per search query
      - constraint:
          # https://ai.google.dev/gemini-api/docs/pricing - "$1.50 starting January 1, 2027"
          start_date: 2027-01-01
        prices:
          input_mtok: 1.5
          cache_read_mtok: 0.15
          output_mtok: 7.5
          web_searches_kcount: 14 # per search query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2027-01-01 }
        prices:
          input_mtok: 0.75
          cache_read_mtok: 0.075
          output_mtok: 3.75
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.375
          cache_read_mtok: 0.0375
          output_mtok: 1.875

  - id: gemini-3.7-flash
    name: Gemini 3.7 Flash
    description: >-
      Google's most capable Flash model, built for complex coding, web development, agentic workflows
      and reliable multi-step execution.
    match:
      starts_with: gemini-3.7-flash
    context_window: 1000000
    prices_checked: 2026-09-29
    price_comments: See https://ai.google.dev/gemini-api/docs/pricing. Standard tier pricing shown; Batch and Flex
      tiers offer 50% discount, Priority tier is 1.8x. Introductory rates apply through 2026-12-31. The cache
      storage price ($0.50 per 1M tokens per hour) has no unit in prices/units.yml and is not represented here.
    prices:
      - prices:
          input_mtok: 0.75
          cache_read_mtok: 0.075
          output_mtok: 3.75
          web_searches_kcount: 14 # per search query
      - constraint:
          # https://ai.google.dev/gemini-api/docs/pricing - "$1.50 starting January 1, 2027"
          start_date: 2027-01-01
        prices:
          input_mtok: 1.5
          cache_read_mtok: 0.15
          output_mtok: 7.5
          web_searches_kcount: 14 # per search query

  - id: gemini-3.8-flash
    name: Gemini 3.8 Flash
    description: >-
      Google's most intelligent Flash model, engineered for long-horizon software engineering,
      autonomous agents, and complex enterprise workflows.
    match:
      or:
        # Excludes the token-priced TTS variants, which are separate models priced below.
        - and:
            - starts_with: gemini-3.8-flash
            - regex: "^(?!.*-[tT][tT][sS](?:$|-))"
        # Family alias: the Gemini API's `modelVersion` resolved it to this model on 2026-09-29.
        # Move it when Google points it at a newer model.
        - equals: gemini-flash-latest
    context_window: 1000000
    prices_checked: 2026-10-14
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Standard tier pricing shown; Batch and Flex tiers offer 50%
      discount, Priority tier is 1.8x. Introductory rates apply through 2026-12-31. The cache storage price ($0.50 per
      1M tokens per hour) has no unit in prices/units.yml and is not represented here. Batch rates verified on
      2026-10-14 at https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is not
      modelled. Batch halves input, output and context caching, at both the introductory and the 2027-01-01 rates.
    prices:
      - prices:
          input_mtok: 0.75
          cache_read_mtok: 0.075
          output_mtok: 3.75
          web_searches_kcount: 14 # per search query
      - constraint:
          # https://ai.google.dev/gemini-api/docs/pricing - "$1.50 starting January 1, 2027"
          start_date: 2027-01-01
        prices:
          input_mtok: 1.5
          cache_read_mtok: 0.15
          output_mtok: 7.5
          web_searches_kcount: 14 # per search query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2027-01-01 }
        prices:
          input_mtok: 0.75
          cache_read_mtok: 0.075
          output_mtok: 3.75
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.375
          cache_read_mtok: 0.0375
          output_mtok: 1.875

  - id: gemini-3.8-flash-lite-tts
    name: Gemini 3.8 Flash-Lite TTS
    description: Token-priced text-to-speech model, the lightest member of the Gemini 3.8 family.
    match:
      equals: gemini-3.8-flash-lite-tts
    prices_checked: 2026-10-15
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Token-priced TTS rates; introductory rates apply through
      2026-12-31. Batch and Flex halve input and output; Priority is 1.8x standard at both rate levels.
    prices:
      - prices:
          input_mtok: 0.5
          output_mtok: 6
      - constraint:
          # https://ai.google.dev/gemini-api/docs/pricing - "$1.00 starting January 1, 2027"
          start_date: 2027-01-01
        prices:
          input_mtok: 1
          output_mtok: 12
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2027-01-01 }
        prices:
          input_mtok: 0.5
          output_mtok: 6
      - when: { processing: batch }
        constraint: { start_date: 2026-10-15 }
        prices:
          input_mtok: 0.25
          output_mtok: 3
      - when: { service_tier: flex }
        constraint: { start_date: 2027-01-01 }
        prices:
          input_mtok: 0.5
          output_mtok: 6
      - when: { service_tier: flex }
        constraint: { start_date: 2026-10-15 }
        prices:
          input_mtok: 0.25
          output_mtok: 3
      - when: { service_tier: priority }
        constraint: { start_date: 2027-01-01 }
        prices:
          input_mtok: 1.8
          output_mtok: 21.6
      - when: { service_tier: priority }
        constraint: { start_date: 2026-10-15 }
        prices:
          input_mtok: 0.9
          output_mtok: 10.8

  - id: gemini-3.8-flash-tts
    name: Gemini 3.8 Flash TTS
    description: Token-priced text-to-speech model in the Gemini 3.8 Flash family.
    match:
      equals: gemini-3.8-flash-tts
    prices_checked: 2026-10-15
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Token-priced TTS rates; introductory rates apply through
      2026-12-31. Batch and Flex halve input and output; Priority is 1.8x standard at both rate levels.
    prices:
      - prices:
          input_mtok: 0.5
          output_mtok: 9
      - constraint:
          # https://ai.google.dev/gemini-api/docs/pricing - "$1.00 starting January 1, 2027"
          start_date: 2027-01-01
        prices:
          input_mtok: 1
          output_mtok: 18
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2027-01-01 }
        prices:
          input_mtok: 0.5
          output_mtok: 9
      - when: { processing: batch }
        constraint: { start_date: 2026-10-15 }
        prices:
          input_mtok: 0.25
          output_mtok: 4.5
      - when: { service_tier: flex }
        constraint: { start_date: 2027-01-01 }
        prices:
          input_mtok: 0.5
          output_mtok: 9
      - when: { service_tier: flex }
        constraint: { start_date: 2026-10-15 }
        prices:
          input_mtok: 0.25
          output_mtok: 4.5
      - when: { service_tier: priority }
        constraint: { start_date: 2027-01-01 }
        prices:
          input_mtok: 1.8
          output_mtok: 32.4
      - when: { service_tier: priority }
        constraint: { start_date: 2026-10-15 }
        prices:
          input_mtok: 0.9
          output_mtok: 16.2

  - id: gemini-3.8-live
    name: Gemini 3.8 Live
    description: Google's low-latency audio-to-audio Live API models for real-time voice agents and live dialogue,
      including Gemini 3.8 Live Extended Thinking, which reasons in the background while it keeps talking.
    match:
      # Also covers `gemini-3.8-live-extended-thinking`, which Google prices identically.
      starts_with: gemini-3.8-live
    context_window: 131072
    prices_checked: 2026-09-29
    price_comments: See https://ai.google.dev/gemini-api/docs/pricing - Gemini 3.8 Live and Gemini 3.8 Live Extended
      Thinking are priced the same as Gemini 3.1 Flash Live Preview, with thinking tokens included in the output
      price. Audio and image/video are also offered per minute, which has no unit here; the per-token prices are shown.
    prices:
      input_mtok: 0.75
      input_audio_mtok: 3
      input_image_mtok: 1
      input_video_mtok: 1
      output_mtok: 4.5
      output_audio_mtok: 12
      web_searches_kcount: 14 # per search query

  - id: gemini-4-argon
    name: Gemini 4 Argon
    description: >-
      Google's frontier model for complex, long-horizon workflows across real-world software engineering,
      enterprise knowledge work like legal and finance, and cybersecurity defense, with output of up to 1M tokens.
    match:
      or:
        - equals: gemini-4-argon
        - starts_with: gemini-4-argon-preview
        - regex: '^gemini-4-argon-\d'
    context_window: 1048576
    prices_checked: 2026-10-14
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Standard tier pricing shown; Batch and Flex tiers offer 50%
      discount, Priority tier is 1.8x. Introductory rates apply through 2026-11-30. No prompt-size tiers or separate
      audio input rate are documented. Implicit caching only. Batch rates verified on 2026-10-14 at
      https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is not modelled. Batch halves
      input, output and context caching, at both the introductory and the 2026-12-01 rates.
    prices:
      - prices:
          input_mtok: 2
          cache_read_mtok: 0.1
          output_mtok: 10
          web_searches_kcount: 14 # per search query
      - constraint:
          # https://ai.google.dev/gemini-api/docs/pricing - "$4.00 starting December 1, 2026"
          start_date: 2026-12-01
        prices:
          input_mtok: 4
          cache_read_mtok: 0.2
          output_mtok: 20
          web_searches_kcount: 14 # per search query
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-12-01 }
        prices:
          input_mtok: 2
          cache_read_mtok: 0.1
          output_mtok: 10
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 1
          cache_read_mtok: 0.05
          output_mtok: 5

  - id: gemini-embedding-001
    match:
      equals: gemini-embedding-001
    prices:
      input_mtok: 0.15

  - id: gemini-embedding-2
    name: Gemini Embedding 2
    description: Google's multimodal embedding model for text, images, audio, video, and documents.
    match:
      or:
        - equals: gemini-embedding-2
        - equals: gemini-embedding-2-preview
    context_window: 8192
    prices_checked: 2026-10-15
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing#gemini-embedding-2. Batch rates verified on 2026-10-14 at
      https://ai.google.dev/gemini-api/docs/pricing (Batch tab); batch history before then is not modelled. Batch halves
      every input modality. The models catalog (https://ai.google.dev/gemini-api/docs/models) names the endpoint
      `gemini-embedding-2-preview` and the pricing page `gemini-embedding-2`; both were added to the match on
      2026-10-15 at the published Embedding 2 rates.
    prices:
      input_mtok: 0.2
      input_image_mtok: 0.45
      input_audio_mtok: 6.5
      input_video_mtok: 12
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.1
          input_image_mtok: 0.225
          input_audio_mtok: 3.25
          input_video_mtok: 6

  - id: gemini-flash-1.5
    name: Gemini 1.5 Flash
    description: >-
      Gemini 1.5 Flash is a foundation model that performs well at a variety of multimodal tasks such as visual understanding,
      classification, summarization, and creating content from image, audio and video.
      It's adept at processing visual and text inputs such as photographs, documents, infographics, and screenshots.
    match:
      equals: gemini-flash-1.5
    prices_checked: 2025-07-05
    price_comments: See https://ai.google.dev/gemini-api/docs/pricing#gemini-1.5-flash
    prices:
      input_mtok:
        base: 0.075
        tiers:
          - start: 128000
            price: 0.15
      cache_read_mtok:
        base: 0.01875
        tiers:
          - start: 128000
            price: 0.0375
      output_mtok:
        base: 0.3
        tiers:
          - start: 128000
            price: 0.6

  - id: gemini-flash-1.5-8b
    name: gemini 1.5 flash
    description: >-
      A faster, more cost-effective variant of Gemini 1.5 that maintains strong
      capabilities while optimizing for performance and cost efficiency.
      Suitable for production deployments requiring high throughput.
    match:
      equals: gemini-flash-1.5-8b
    context_window: 1000000
    prices_checked: 2025-07-05
    price_comments: See https://ai.google.dev/gemini-api/docs/pricing#gemini-1.5-flash-8b
    prices:
      input_mtok:
        base: 0.0375
        tiers:
          - start: 128000
            price: 0.075
      cache_read_mtok:
        base: 0.01
        tiers:
          - start: 128000
            price: 0.02
      output_mtok:
        base: 0.15
        tiers:
          - start: 128000
            price: 0.3

  - id: gemini-live-2.5-flash
    name: Gemini Live 2.5 Flash
    description: >-
      Google's Live API model for low-latency bidirectional voice and video
      interactions, GA on Vertex AI (model id `gemini-live-2.5-flash`, served
      from the `global` location). The prefix match also covers the AI Studio
      preview ids (`gemini-live-2.5-flash-preview*`).
    match:
      or:
        - starts_with: gemini-live-2.5-flash
        - starts_with: gemini-2.5-flash-native-audio-preview
        - equals: gemini-2.5-flash-native-audio-latest
    price_comments: >-
      See https://cloud.google.com/vertex-ai/generative-ai/pricing (Live API)
      and https://ai.google.dev/gemini-api/docs/pricing - GA pricing matches
      the preview.
    prices_checked: 2026-07-24
    prices:
      input_mtok: 0.5
      input_audio_mtok: 3
      output_mtok: 2
      output_audio_mtok: 12

  - id: gemini-nano-banana-2.1
    name: Gemini Nano Banana 2.1
    description: >-
      Google's latest image generation and editing model, with text/thinking output tokens priced
      separately from image output tokens.
    match:
      equals: gemini-nano-banana-2.1
    prices_checked: 2026-10-15
    price_comments: >-
      See https://ai.google.dev/gemini-api/docs/pricing. Image output is $30 per 1M tokens; text and thinking
      output is $7.50 per 1M tokens and is the aggregate remainder when modality details omit text.
      Batch halves input, text output and image output. Google Web and Image Search grounding is $14 per
      1,000 billable search queries; callers supply the count after account-wide free allowances.
    prices:
      input_mtok: 1.5
      output_mtok: 7.5
      output_image_mtok: 30
      web_searches_kcount: 14
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-15 }
        prices:
          input_mtok: 0.75
          output_mtok: 3.75
          output_image_mtok: 15

  - id: gemini-omni-1.1-flash
    name: Gemini Omni Flash
    description: Google's video generation and editing model with multimodal input.
    match:
      equals: gemini-omni-1.1-flash
    prices_checked: 2026-10-15
    price_comments: >-
      Rates from https://ai.google.dev/gemini-api/docs/pricing (Gemini Omni Flash, GA): input
      $1.50 / 1M tokens across text, image, video and audio; output $9 / 1M tokens (text) and
      $17.50 / 1M tokens (video). The models catalog names `gemini-omni-1.1-flash` as the GA endpoint of
      the Omni Flash family.
    prices:
      input_mtok: 1.5
      output_mtok: 9
      output_video_mtok: 17.5

  - id: gemini-omni-flash-preview
    name: Gemini Omni Flash Preview
    description: Google's video generation and editing model with multimodal input.
    match:
      starts_with: gemini-omni-flash-preview
    prices_checked: 2026-07-22
    price_comments: See https://ai.google.dev/gemini-api/docs/pricing#gemini-omni-flash-preview.
    prices:
      input_mtok: 1.5
      output_mtok: 9
      output_video_mtok: 17.5

  - id: gemini-pro
    name: gemini 1.0 pro
    description: >-
      Google's first-generation advanced multimodal model that can understand
      text, code, and images. It provides strong reasoning capabilities and
      follows instructions effectively.
    match:
      or:
        - equals: gemini-pro
        - equals: gemini-1.0-pro
    context_window: 32768
    prices_checked: 2025-07-04
    price_comments: I can't find anything so trusting these prices, not sure the model still exists
    prices:
      input_mtok: 0.125
      output_mtok: 0.375

  - id: gemini-pro-1.5
    name: Gemini 1.5 Pro
    description: >-
      Google's latest multimodal model, supports image and video[0] in text or chat prompts.
    match:
      equals: gemini-pro-1.5
    context_window: 2000000
    prices_checked: 2025-07-05
    price_comments: See https://ai.google.dev/gemini-api/docs/pricing#gemini-1.5-pro
    prices:
      input_mtok:
        base: 1.25
        tiers:
          - start: 128000
            price: 2.5
      cache_read_mtok:
        base: 0.3125
        tiers:
          - start: 128000
            price: 0.625
      output_mtok:
        base: 5
        tiers:
          - start: 128000
            price: 10

  - id: gemini-robotics-er-2-preview
    match:
      equals: gemini-robotics-er-2-preview
    prices_checked: 2026-10-15
    price_comments: >-
      Rates from https://ai.google.dev/gemini-api/docs/pricing (Gemini Robotics ER 2): input $1 / 1M tokens
      and output $5 / 1M tokens through 2026-12-31, rising to $2 and $10 from 2027-01-01. Cache reads are
      implicit only at $0.10 / 1M tokens through 2026-12-31 and $0.20 from 2027-01-01; the engines do not
      represent the cache storage fee.
    prices:
      - prices:
          input_mtok: 1
          cache_read_mtok: 0.1
          output_mtok: 5
      - constraint:
          # https://ai.google.dev/gemini-api/docs/pricing - "$2.00 starting January 1, 2027"
          start_date: 2027-01-01
        prices:
          input_mtok: 2
          cache_read_mtok: 0.2
          output_mtok: 10

  - id: gemini-robotics-er-2-streaming-preview
    match:
      equals: gemini-robotics-er-2-streaming-preview
    prices_checked: 2026-10-15
    price_comments: >-
      Rates from https://ai.google.dev/gemini-api/docs/pricing (Gemini Robotics ER 2 Streaming): input
      $1 / 1M tokens and output $5 / 1M tokens through 2026-12-31, rising to $2 and $10 from 2027-01-01.
      No context-caching section is published for the streaming endpoint.
    prices:
      - prices:
          input_mtok: 1
          output_mtok: 5
      - constraint:
          # https://ai.google.dev/gemini-api/docs/pricing - "$2.00 starting January 1, 2027"
          start_date: 2027-01-01
        prices:
          input_mtok: 2
          output_mtok: 10

  - id: gemma-2-27b-it
    name: Gemma 2 27B
    description: >-
      Gemma 2 27B by Google is an open model built from the same research and technology used to create the Gemini models.
      Gemma models are well-suited for a variety of text generation and instruction-following tasks.
    match:
      equals: gemma-2-27b-it
    prices_checked: 2026-06-09
    price_comments: Imported from OpenRouter pricing; verify against Google pricing when native API pricing is published.
    prices:
      input_mtok: 0.65
      output_mtok: 0.65

  - id: gemma-3
    name: Gemma 3 (free)
    description: Lightweight, state-of the art, open model built from the same technology that powers our Gemini models.
    prices_checked: 2025-08-12
    match:
      or:
        - starts_with: gemma-3-
        - equals: gemma-3
    prices: {}

  - id: gemma-3n
    name: Gemma 3n (free)
    description: Our open model built for efficient performance on everyday devices like mobile phones, laptops, and
      tablets.
    collapse: false
    prices_checked: 2025-08-12
    match:
      or:
        - starts_with: gemma-3n
    prices: {}

  - id: gemma-4-26b-a4b-it
    name: Gemma 4 26B A4B
    description: >-
      Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total
      parameters, only 3.8B activate per token during inference.
    match:
      equals: gemma-4-26b-a4b-it
    context_window: 262144
    prices_checked: 2026-08-27
    price_comments: >-
      Gemini API model ID. Free of charge on the Gemini API with no paid tier, see
      https://ai.google.dev/gemini-api/docs/pricing. The Vertex AI Model-as-a-Service offering uses the
      gemma-4-26b-a4b-it-maas ID and is priced separately. The previous rates were imported from OpenRouter, not Google.
    prices: {}

  - id: gemma-4-26b-a4b-it-maas
    name: Gemma 4 26B A4B (Vertex AI MaaS)
    description: >-
      Gemma 4 26B A4B IT served by Vertex AI as a managed Model-as-a-Service endpoint.
    match:
      equals: gemma-4-26b-a4b-it-maas
    context_window: 262144
    prices_checked: 2026-08-27
    price_comments: >-
      Vertex AI Model-as-a-Service rates, see https://cloud.google.com/vertex-ai/generative-ai/pricing. Model ID from
      https://cloud.google.com/vertex-ai/generative-ai/docs/maas/use-open-models.
    prices:
      input_mtok: 0.15
      cache_read_mtok: 0.015
      output_mtok: 0.6

  - id: gemma-4-31b-it
    name: Gemma 4 31B
    description: >-
      Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output.
      It features a 256K token context window, configurable thinking/reasoning mode, and native function calling.
    match:
      equals: gemma-4-31b-it
    context_window: 262144
    prices_checked: 2026-08-27
    price_comments: >-
      Free of charge on the Gemini API with no paid tier, see https://ai.google.dev/gemini-api/docs/pricing. Not offered
      as Model-as-a-Service on Vertex AI (only Gemma 4 26B is); there it is self-deployed from Model Garden and billed as
      endpoint compute, so no per-token price applies. The previous rates were imported from OpenRouter, not Google.
    prices: {}