{
  "_meta": {
    "source": "https://apipriceindex.com/api/groq/gpt-oss-20b",
    "attribution": "apipriceindex.com",
    "license": "CC BY 4.0",
    "license_url": "https://creativecommons.org/licenses/by/4.0/",
    "note": "Free to reuse with attribution to apipriceindex.com.",
    "generated_at": "2026-09-05T08:15:05.851Z"
  },
  "id": "groq:gpt-oss-20b",
  "name": "GPT-OSS 20B (Groq)",
  "base_model": "gpt-oss-20b",
  "provider": {
    "id": "groq",
    "name": "Groq",
    "url": "https://groq.com"
  },
  "category": "llm-chat",
  "capabilities": [
    "text",
    "function-calling"
  ],
  "context_window": 128000,
  "max_output": 32768,
  "pricing": {
    "model_id": "groq:gpt-oss-20b",
    "input": 0.075,
    "output": 0.3,
    "cached_input": 0.0375,
    "source_url": "https://openrouter.ai/openai/gpt-oss-20b",
    "verified_at": "2026-09-05T08:00:57Z",
    "sources": [
      {
        "name": "OpenRouter",
        "input": 0.075,
        "output": 0.3,
        "url": "https://openrouter.ai/openai/gpt-oss-20b",
        "verified_at": "2026-09-05T08:00:57Z"
      },
      {
        "name": "LiteLLM",
        "input": 0.075,
        "output": 0.3,
        "url": "https://github.com/BerriAI/litellm/blob/main/model_prices_and_context_window.json",
        "verified_at": "2026-09-03T20:47:46Z"
      },
      {
        "name": "models.dev",
        "input": 0.075,
        "output": 0.3,
        "url": "https://models.dev/",
        "verified_at": "2026-09-04T10:52:49Z"
      }
    ],
    "confidence": "corroborated",
    "delta_pct": 0
  },
  "price_history": [
    [
      "2026-08-05",
      0.075,
      0.3
    ],
    [
      "2026-09-01",
      0.075,
      0.3
    ]
  ],
  "latency": [
    {
      "model_id": "groq:gpt-oss-20b",
      "region": "eu-frankfurt",
      "ttft_ms": {
        "p50": 390.3,
        "p95": 475.7
      },
      "total_ms": {
        "p50": 532,
        "p95": 607
      },
      "tokens_per_s": 1869.7,
      "samples": 5,
      "error_rate": 0,
      "token_source": "usage",
      "measured_at": "2026-09-04T18:43:08Z"
    },
    {
      "model_id": "groq:gpt-oss-20b",
      "region": "eu-paris",
      "ttft_ms": {
        "p50": 345.3,
        "p95": 419.2
      },
      "total_ms": {
        "p50": 490.8,
        "p95": 529.8
      },
      "tokens_per_s": 1987,
      "samples": 5,
      "error_rate": 0,
      "token_source": "usage",
      "measured_at": "2026-09-04T18:44:12Z"
    }
  ],
  "license": {
    "license": "Apache 2.0",
    "tier": "permissive",
    "open_weights": true,
    "source_url": "https://huggingface.co/openai/gpt-oss-120b"
  },
  "cutoff": {
    "cutoff": "2024-06",
    "label": "June 2024",
    "source_url": "https://platform.openai.com/docs/models"
  },
  "intelligence": {
    "score": 15.2,
    "model_name": "GPT-OSS 20B",
    "verified_at": "2026-09-04",
    "benchmark_label": "Artificial Analysis Intelligence Index",
    "source_url": "https://artificialanalysis.ai/"
  },
  "quality": {
    "elo": 1317,
    "arena_name": "gpt-oss-20b",
    "verified_at": "2026-09-04",
    "benchmark_label": "LMArena (Chatbot Arena)",
    "source_url": "https://lmarena.ai/leaderboard/text"
  },
  "api": {
    "base_url": "https://api.groq.com/openai/v1",
    "model": "openai/gpt-oss-20b",
    "via": "first-party",
    "key_env": "GROQ_API_KEY",
    "docs_url": null
  },
  "facts": [
    "GPT-OSS 20B (Groq) costs $0.07 input / $0.30 output per 1M tokens.",
    "Cached input cuts the input price by 50%.",
    "Output price has held steady at $0.30 per 1M tokens since 2026-08-05.",
    "A chatbot workload (1000 in / 500 out tokens) costs $0.23 per 1,000 requests.",
    "Cross-checked: LiteLLM's public price index independently lists Groq's gpt-oss-20b at $0.07/$0.30 per 1M — corroborating the price shown here (within 0%).",
    "Handles a 128,000-token context window, up to 32,768 output tokens.",
    "Open weights under Apache 2.0 — permissive licence.",
    "Training knowledge cutoff: June 2024.",
    "Scores 1317 on the LMArena (Chatbot Arena) (human-preference leaderboard) for gpt-oss-20b.",
    "Scores 15.2 on the Artificial Analysis Intelligence Index for gpt-oss-20b.",
    "From eu-frankfurt, p50 time-to-first-token is 390 ms (5 samples).",
    "From eu-paris, p50 time-to-first-token is 345 ms (5 samples).",
    "Groq can be tried with no credit card.",
    "212% pricier than the cheapest gpt-oss-20b endpoint ($0.12 on AkashML).",
    "Value: 1317 Elo at $0.38 blended = 3512 Elo points per dollar — ranking #26 of 321 priced & rated endpoints in the index.",
    "Intelligence value: 15.2 index at $0.38 blended = 40.5 index points per dollar — ranking #79 of 373 priced & AA-rated endpoints in the index.",
    "Throughput: 1987 tokens/s measured from eu-paris at $0.38 blended = 5298.7 tokens/s per dollar — ranking #1 of 11 priced & speed-tested endpoints in the index."
  ]
}