groq.yml

# yaml-language-server: $schema=.schema.json
name: Groq
id: groq
pricing_urls:
  - https://console.groq.com/docs/models
api_pattern: 'https://api\.groq\.com'

# https://github.com/groq/groq-python/blob/v0.31.0/src/groq/types/completion_usage.py
extractors:
  - root: usage
    mappings:
      - path: prompt_tokens
        dest: input_tokens
      - path: completion_tokens
        dest: output_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

models:
  - id: canopylabs/orpheus-arabic-saudi
    name: Orpheus Arabic Saudi (Canopy Labs)
    match:
      equals: canopylabs/orpheus-arabic-saudi
    context_window: 4000
    prices_checked: 2026-10-10
    price_comments: "https://console.groq.com/docs/models#canopylabs/orpheus-arabic-saudi"
    prices:
      input_mchars: 40

  - id: canopylabs/orpheus-v1-english
    name: Orpheus V1 English (Canopy Labs)
    match:
      equals: canopylabs/orpheus-v1-english
    context_window: 4000
    prices_checked: 2026-10-10
    price_comments: "https://console.groq.com/docs/models#canopylabs/orpheus-v1-english"
    prices:
      input_mchars: 22

  - id: deepseek-r1-distill-llama-70b
    name: DeepSeek R1 Distill Llama 70B
    context_window: 131072
    match:
      equals: deepseek-r1-distill-llama-70b
    prices_checked: 2025-08-12
    prices:
      input_mtok: 0.75
      output_mtok: 0.99

  - id: gemma-7b-it
    match:
      equals: gemma-7b-it
    prices:
      input_mtok: 0.07
      output_mtok: 0.07

  - id: gemma2-9b-it
    name: Gemma 2 9B 8k
    match:
      or:
        - equals: gemma2-9b-it
        - equals: gemma2-9b
    prices_checked: 2025-08-12
    prices:
      input_mtok: 0.2
      output_mtok: 0.2

  - id: llama-3.1-405b-reasoning
    match:
      equals: llama-3.1-405b-reasoning
    prices:
      input_mtok: 0.59
      output_mtok: 0.79

  - id: llama-3.1-70b-versatile
    match:
      equals: llama-3.1-70b-versatile
    prices:
      input_mtok: 0.59
      output_mtok: 0.79

  - id: llama-3.1-8b-instant
    name: Llama 3.1 8B Instant 128k
    match:
      equals: llama-3.1-8b-instant
    prices_checked: 2025-08-12
    prices:
      input_mtok: 0.05
      output_mtok: 0.08

  - id: llama-3.2-11b-text-preview
    match:
      equals: llama-3.2-11b-text-preview
    prices:
      input_mtok: 0.18
      output_mtok: 0.18

  - id: llama-3.2-11b-vision-preview
    match:
      equals: llama-3.2-11b-vision-preview
    prices:
      input_mtok: 0.18
      output_mtok: 0.18

  - id: llama-3.2-1b-preview
    match:
      equals: llama-3.2-1b-preview
    prices:
      input_mtok: 0.04
      output_mtok: 0.04

  - id: llama-3.2-3b-preview
    match:
      equals: llama-3.2-3b-preview
    prices:
      input_mtok: 0.06
      output_mtok: 0.06

  - id: llama-3.2-90b-text-preview
    match:
      equals: llama-3.2-90b-text-preview
    prices:
      input_mtok: 0.9
      output_mtok: 0.9

  - id: llama-3.2-90b-vision-preview
    match:
      equals: llama-3.2-90b-vision-preview
    prices:
      input_mtok: 0.9
      output_mtok: 0.9

  - id: llama-3.3-70b-specdec
    match:
      equals: llama-3.3-70b-specdec
    prices:
      input_mtok: 0.59
      output_mtok: 0.99

  - id: llama-3.3-70b-versatile
    name: Llama 3.3 70B Versatile 128k
    match:
      equals: llama-3.3-70b-versatile
    prices_checked: 2025-08-12
    prices:
      input_mtok: 0.59
      output_mtok: 0.79

  - id: llama-guard-3-8b
    match:
      equals: llama-guard-3-8b
    prices:
      input_mtok: 0.2
      output_mtok: 0.2

  - id: llama2-70b-4096
    match:
      equals: llama2-70b-4096
    prices:
      input_mtok: 0.7
      output_mtok: 0.8

  - id: llama3-70b-8192
    match:
      equals: llama3-70b-8192
    prices:
      input_mtok: 0.59
      output_mtok: 0.79

  - id: llama3-8b-8192
    match:
      equals: llama3-8b-8192
    prices:
      input_mtok: 0.05
      output_mtok: 0.08

  - id: llama3-groq-70b-8192-tool-use-preview
    match:
      equals: llama3-groq-70b-8192-tool-use-preview
    prices:
      input_mtok: 0.89
      output_mtok: 0.89

  - id: llama3-groq-8b-8192-tool-use-preview
    match:
      equals: llama3-groq-8b-8192-tool-use-preview
    prices:
      input_mtok: 0.19
      output_mtok: 0.19

  - id: meta-llama/llama-4-maverick-17b-128e-instruct
    name: Llama 4 Maverick 17B 128E
    context_window: 131072
    match:
      equals: meta-llama/llama-4-maverick-17b-128e-instruct
    prices_checked: 2025-08-12
    prices:
      input_mtok: 0.2
      output_mtok: 0.6

  - id: meta-llama/llama-4-scout-17b-16e-instruct
    name: Llama 4 Scout (17Bx16E) 128k
    match:
      equals: meta-llama/llama-4-scout-17b-16e-instruct
    prices_checked: 2025-08-12
    prices:
      input_mtok: 0.11
      output_mtok: 0.34

  - id: meta-llama/llama-guard-4-12b
    name: Llama Guard 4 12B
    context_window: 131072
    match:
      equals: meta-llama/llama-guard-4-12b
    prices_checked: 2025-08-12
    prices:
      input_mtok: 0.2
      output_mtok: 0.2

  - id: meta-llama/llama-prompt-guard-2-22m
    name: Llama Prompt Guard 2 22M
    match:
      equals: meta-llama/llama-prompt-guard-2-22m
    context_window: 512
    prices_checked: 2026-10-10
    price_comments: "https://console.groq.com/docs/models#meta-llama/llama-prompt-guard-2-22m"
    prices:
      input_mtok: 0.03
      output_mtok: 0.03

  - id: meta-llama/llama-prompt-guard-2-86m
    name: Llama Prompt Guard 2 86M
    match:
      equals: meta-llama/llama-prompt-guard-2-86m
    context_window: 512
    prices_checked: 2026-10-10
    price_comments: "https://console.groq.com/docs/models#meta-llama/llama-prompt-guard-2-86m"
    prices:
      input_mtok: 0.04
      output_mtok: 0.04

  - id: mistral-saba-24b
    match:
      equals: mistral-saba-24b
    prices:
      input_mtok: 0.79
      output_mtok: 0.79

  - id: mixtral-8x7b-32768
    match:
      equals: mixtral-8x7b-32768
    prices:
      input_mtok: 0.24
      output_mtok: 0.24

  - id: moonshotai/kimi-k2-instruct
    name: Kimi K2 1T 128k
    context_window: 131072
    match:
      or:
        - equals: moonshotai/kimi-k2-instruct
        - equals: moonshotai/kimi-k2-instruct-0905
    prices_checked: 2025-08-12
    prices:
      input_mtok: 1
      cache_read_mtok: 0.5
      output_mtok: 3

  # https://console.groq.com/docs/model/openai/gpt-oss-120b
  - id: openai/gpt-oss-120b
    description: |
      GPT-OSS 120B is OpenAI's flagship open source model, built on a Mixture-of-Experts (MoE) architecture with
      120 billion parameters and 128 experts.
    match:
      equals: openai/gpt-oss-120b
    context_window: 131072
    prices_checked: 2026-10-14
    price_comments: >-
      Batch API is 50% of the synchronous rates and does not stack with prompt caching: cached input is billed
      at the halved input rate. See https://console.groq.com/docs/batch.
    prices:
      input_mtok: 0.15
      cache_read_mtok: 0.075
      output_mtok: 0.6
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.075
          cache_read_mtok: 0.075
          output_mtok: 0.3

  # https://console.groq.com/docs/model/openai/gpt-oss-20b
  - id: openai/gpt-oss-20b
    description: |
      GPT-OSS 20B is OpenAI's flagship open source model, built on a Mixture-of-Experts (MoE) architecture with
      20 billion parameters and 32 experts.
    match:
      equals: openai/gpt-oss-20b
    context_window: 131072
    prices_checked: 2026-10-14
    price_comments: >-
      Batch API is 50% of the synchronous rates and does not stack with prompt caching: cached input is billed
      at the halved input rate. See https://console.groq.com/docs/batch.
    prices:
      input_mtok: 0.075
      cache_read_mtok: 0.0375
      output_mtok: 0.30
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          input_mtok: 0.0375
          cache_read_mtok: 0.0375
          output_mtok: 0.15

  # https://console.groq.com/docs/model/openai/gpt-oss-safeguard-20b
  - id: openai/gpt-oss-safeguard-20b
    name: GPT-OSS Safeguard 20B
    description: |
      GPT-OSS Safeguard 20B is OpenAI's open-weight safety model. It classifies content against a policy supplied
      in the prompt.
    match:
      equals: openai/gpt-oss-safeguard-20b
    context_window: 131072
    prices_checked: 2026-09-29
    prices:
      input_mtok: 0.075
      cache_read_mtok: 0.0375
      output_mtok: 0.30

  - id: qwen/qwen3-32b
    name: Qwen3 32B 131k
    match:
      equals: qwen/qwen3-32b
    context_window: 131072
    prices_checked: 2025-08-12
    prices:
      input_mtok: 0.29
      output_mtok: 0.59

  # https://console.groq.com/docs/model/qwen/qwen3.8-27b
  - id: qwen/qwen3.8-27b
    name: Qwen 3.8 27B
    match:
      equals: qwen/qwen3.8-27b
    context_window: 131072
    prices_checked: 2026-09-29
    price_comments: Preview model. Groq lists no cached-input rate.
    prices:
      input_mtok: 0.80
      output_mtok: 4.00

  - id: whisper-large-v3
    name: Whisper Large V3
    match:
      equals: whisper-large-v3
    # No context_window: Groq's catalog reports 448 here, but that is the audio decoder's
    # token limit, not a text context window
    prices_checked: 2026-10-14
    price_comments: >-
      See https://console.groq.com/docs/model/whisper-large-v3. Batch API is 50% of the synchronous rate, see
      https://console.groq.com/docs/batch.
    prices:
      audio_hours: 0.111
      input_audio_hours: 0.111
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          audio_hours: 0.0555
          input_audio_hours: 0.0555

  - id: whisper-large-v3-turbo
    name: Whisper Large V3 Turbo
    match:
      equals: whisper-large-v3-turbo
    # No context_window: Groq's catalog reports 448 here, but that is the audio decoder's
    # token limit, not a text context window
    prices_checked: 2026-10-14
    price_comments: >-
      See https://console.groq.com/docs/model/whisper-large-v3-turbo. Batch API is 50% of the synchronous rate, see
      https://console.groq.com/docs/batch.
    prices:
      audio_hours: 0.04
      input_audio_hours: 0.04
    price_variants:
      - when: { processing: batch }
        constraint: { start_date: 2026-10-14 }
        prices:
          audio_hours: 0.02
          input_audio_hours: 0.02