{
  "source": "https://healthbenchhard.ai",
  "license": "CC BY 4.0",
  "citation": "Arcophos. HealthBench Hard Leaderboard. 2026-08-16 snapshot. https://healthbenchhard.ai.",
  "updated": "2026-08-16",
  "benchmark": {
    "name": "HealthBench Hard",
    "scale": "0 to 1",
    "labs": 2,
    "tasks": 1000,
    "sourceConversations": 5000,
    "physicians": 262,
    "countries": 60,
    "criteria": 48562,
    "grader": "GPT-4.1",
    "launchTopScore": 0.32,
    "paper": "https://arxiv.org/abs/2505.08775",
    "publisher": "OpenAI",
    "released": "2025-05-12",
    "pricingNote": "GPT-5.6 models charge higher rates above 272K input tokens. GPT OSS models are open weights without vendor list pricing.",
    "conversations": 1000
  },
  "models": [
    {
      "rank": 1,
      "slug": "muse-spark",
      "name": "Muse Spark",
      "lab": "Meta",
      "logo": "meta",
      "score": 0.428,
      "size": null,
      "context": "1.0M",
      "contextTokens": 1000000,
      "inputPer1M": 1.25,
      "outputPer1M": 4.25,
      "license": "proprietary",
      "releaseDate": "2026-08-05",
      "positioning": "Meta's frontier multimodal reasoning model, sold through the Meta Model API, with weights announced to open later in 2026."
    },
    {
      "rank": 2,
      "slug": "gpt-5.6-sol",
      "name": "GPT-5.6 Sol",
      "lab": "OpenAI",
      "logo": "openai",
      "score": 0.331,
      "size": null,
      "context": "1.1M",
      "contextTokens": 1100000,
      "inputPer1M": 5,
      "outputPer1M": 30,
      "license": "proprietary",
      "releaseDate": "2026-07-09",
      "positioning": "OpenAI's flagship 5.6-series model, built for its heaviest reasoning and agentic work."
    },
    {
      "rank": 3,
      "slug": "gpt-5.6-terra",
      "name": "GPT-5.6 Terra",
      "lab": "OpenAI",
      "logo": "openai",
      "score": 0.327,
      "size": null,
      "context": "1.1M",
      "contextTokens": 1100000,
      "inputPer1M": 2,
      "outputPer1M": 12,
      "license": "proprietary",
      "releaseDate": "2026-07-09",
      "positioning": "The middle of OpenAI's 5.6 lineup, sitting between Sol and Luna on price and capability."
    },
    {
      "rank": 4,
      "slug": "gpt-5.6-luna",
      "name": "GPT-5.6 Luna",
      "lab": "OpenAI",
      "logo": "openai",
      "score": 0.32,
      "size": null,
      "context": "1.1M",
      "contextTokens": 1100000,
      "inputPer1M": 0.2,
      "outputPer1M": 1.2,
      "license": "proprietary",
      "releaseDate": "2026-07-09",
      "positioning": "OpenAI's high-volume 5.6 tier, cut to $0.20 per million input tokens in the July 30, 2026 repricing."
    },
    {
      "rank": 5,
      "slug": "gpt-oss-120b",
      "name": "GPT OSS 120B",
      "lab": "OpenAI",
      "logo": "openai",
      "score": 0.3,
      "size": "117B",
      "context": "131K",
      "contextTokens": 131072,
      "inputPer1M": null,
      "outputPer1M": null,
      "license": "open",
      "releaseDate": "2025-08-05",
      "positioning": "OpenAI's most capable open-weights model, an Apache 2.0 mixture-of-experts release that fits on a single 80GB GPU."
    },
    {
      "rank": 6,
      "slug": "gpt-5.3-chat",
      "name": "GPT-5.3 Chat",
      "lab": "OpenAI",
      "logo": "openai",
      "score": 0.259,
      "size": null,
      "context": "128K",
      "contextTokens": 128000,
      "inputPer1M": 1.75,
      "outputPer1M": 14,
      "license": "proprietary",
      "releaseDate": "2026-03-05",
      "positioning": "The API alias of the non-reasoning ChatGPT default from spring 2026, now deprecated in favor of GPT-5.6."
    },
    {
      "rank": 7,
      "slug": "gpt-5.5-instant",
      "name": "GPT-5.5 Instant",
      "lab": "OpenAI",
      "logo": "openai",
      "score": 0.229,
      "size": null,
      "context": "400K",
      "contextTokens": 400000,
      "inputPer1M": 5,
      "outputPer1M": 30,
      "license": "proprietary",
      "releaseDate": "2026-05-05",
      "positioning": "The speed-focused variant that took over as ChatGPT's default in May 2026."
    },
    {
      "rank": 8,
      "slug": "gpt-oss-20b",
      "name": "GPT OSS 20B",
      "lab": "OpenAI",
      "logo": "openai",
      "score": 0.108,
      "size": "21B",
      "context": "131K",
      "contextTokens": 131072,
      "inputPer1M": null,
      "outputPer1M": null,
      "license": "open",
      "releaseDate": "2025-08-05",
      "positioning": "The smaller Apache 2.0 open-weights release, sized to run locally in 16GB of memory."
    },
    {
      "rank": 9,
      "slug": "gpt-5",
      "name": "GPT-5",
      "lab": "OpenAI",
      "logo": "openai",
      "score": 0.016,
      "size": null,
      "context": "400K",
      "contextTokens": 400000,
      "inputPer1M": 1.25,
      "outputPer1M": 10,
      "license": "proprietary",
      "releaseDate": "2025-08-07",
      "positioning": "OpenAI's August 2025 flagship, still served at unchanged prices but superseded by the GPT-5.6 series."
    }
  ],
  "pending": []
}