{
  "schema_version": "1.0",
  "dataset": "Benchquill AI model leaderboard",
  "canonical_url": "https://benchquill.com/models.json",
  "source_page": "https://benchquill.com/leaderboard",
  "methodology": "https://benchquill.com/methodology",
  "citation": "Benchquill AI Model Leaderboard, accessed [date], https://benchquill.com/",
  "last_verified": "2026-04-29",
  "generated_from": "app.js",
  "model_count": 45,
  "fields": {
    "rank": "Computed from overall descending at export time.",
    "overall": "Composite 0-100 Benchquill editorial score, not a single official benchmark.",
    "code": "Coding capability composite, 0-100.",
    "reasoning": "Reasoning capability composite, 0-100.",
    "math": "Math capability composite, 0-100.",
    "vision": "Vision/multimodal capability composite, 0-100, null when not scored.",
    "speed_tokens_per_second": "Estimated throughput; verify with your own prompt, provider, region, and load.",
    "blended_price_usd_per_million_tokens": "Benchquill weighted estimate: 25% input price plus 75% output price."
  },
  "caveats": [
    "Scores are editorial composites unless a page or CSV row names a raw benchmark source.",
    "Provider pricing can change; verify against official pricing pages before purchase.",
    "GPT-5 nano is source-checked as an OpenAI pricing-only model but excluded from the ranked table until Benchquill has enough comparable benchmark evidence.",
    "DeepSeek V3.2 remains in the ranked archive as a historical Dec 2025 DeepSeek record; current DeepSeek API pricing comparisons should use V4-Pro and V4-Flash."
  ],
  "models": [
    {
      "last_verified": "2026-04-29",
      "rank": 1,
      "rank_basis": "overall_score_descending",
      "slug": "gpt-5-5",
      "name": "GPT-5.5",
      "provider": "OpenAI",
      "tier": "S",
      "overall": 94.6,
      "score_type": "Benchquill editorial composite",
      "code": 93.4,
      "reasoning": 95.2,
      "math": 95.8,
      "vision": 91.4,
      "speed_tokens_per_second": 138,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 5,
      "output_price_usd_per_million_tokens": 30,
      "blended_price_usd_per_million_tokens": 23.75,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1.05M",
      "record_note": "",
      "release_month": "2026-04",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 2,
      "rank_basis": "overall_score_descending",
      "slug": "claude-opus-4-7",
      "name": "Claude Opus 4.7",
      "provider": "Anthropic",
      "tier": "S",
      "overall": 93.8,
      "score_type": "Benchquill editorial composite",
      "code": 95.8,
      "reasoning": 94.2,
      "math": 91.5,
      "vision": 90.7,
      "speed_tokens_per_second": 86,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 5,
      "output_price_usd_per_million_tokens": 25,
      "blended_price_usd_per_million_tokens": 20,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2026-04",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 3,
      "rank_basis": "overall_score_descending",
      "slug": "gemini-3-pro",
      "name": "Gemini 3.1 Pro Preview",
      "provider": "Google",
      "tier": "S",
      "overall": 92.4,
      "score_type": "Benchquill editorial composite",
      "code": 88.1,
      "reasoning": 91.5,
      "math": 93.7,
      "vision": 94.6,
      "speed_tokens_per_second": 192,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 2,
      "output_price_usd_per_million_tokens": 12,
      "blended_price_usd_per_million_tokens": 9.5,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2026-03",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision",
        "Audio"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 4,
      "rank_basis": "overall_score_descending",
      "slug": "gpt-5",
      "name": "GPT-5",
      "provider": "OpenAI",
      "tier": "S",
      "overall": 91.2,
      "score_type": "Benchquill editorial composite",
      "code": 90.4,
      "reasoning": 92.1,
      "math": 93.6,
      "vision": 88.9,
      "speed_tokens_per_second": 154,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 1.25,
      "output_price_usd_per_million_tokens": 10,
      "blended_price_usd_per_million_tokens": 7.81,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "400k",
      "record_note": "",
      "release_month": "2025-08",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 5,
      "rank_basis": "overall_score_descending",
      "slug": "claude-sonnet-4-6",
      "name": "Claude Sonnet 4.6",
      "provider": "Anthropic",
      "tier": "S",
      "overall": 89.8,
      "score_type": "Benchquill editorial composite",
      "code": 92.4,
      "reasoning": 89.7,
      "math": 87.5,
      "vision": 86.2,
      "speed_tokens_per_second": 162,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 2,
      "output_price_usd_per_million_tokens": 6,
      "blended_price_usd_per_million_tokens": 5,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2026-02",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 6,
      "rank_basis": "overall_score_descending",
      "slug": "o3",
      "name": "o3",
      "provider": "OpenAI",
      "tier": "A",
      "overall": 88.9,
      "score_type": "Benchquill editorial composite",
      "code": 87.1,
      "reasoning": 92.4,
      "math": 94.8,
      "vision": 81.3,
      "speed_tokens_per_second": 78,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 2,
      "output_price_usd_per_million_tokens": 8,
      "blended_price_usd_per_million_tokens": 6.5,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "200k",
      "record_note": "",
      "release_month": "2025-04",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 7,
      "rank_basis": "overall_score_descending",
      "slug": "deepseek-v4-pro",
      "name": "DeepSeek V4-Pro",
      "provider": "DeepSeek",
      "tier": "A",
      "overall": 87.9,
      "score_type": "Benchquill editorial composite",
      "code": 80.6,
      "reasoning": 88.4,
      "math": 89.2,
      "vision": null,
      "speed_tokens_per_second": 96,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.435,
      "output_price_usd_per_million_tokens": 0.87,
      "blended_price_usd_per_million_tokens": 0.76,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2026-04",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 8,
      "rank_basis": "overall_score_descending",
      "slug": "gemini-2-5-pro",
      "name": "Gemini 2.5 Pro",
      "provider": "Google",
      "tier": "A",
      "overall": 87.6,
      "score_type": "Benchquill editorial composite",
      "code": 84.2,
      "reasoning": 88.4,
      "math": 90.1,
      "vision": 88.7,
      "speed_tokens_per_second": 154,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 1.25,
      "output_price_usd_per_million_tokens": 10,
      "blended_price_usd_per_million_tokens": 7.81,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2025-03",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 9,
      "rank_basis": "overall_score_descending",
      "slug": "claude-opus-4",
      "name": "Claude Opus 4",
      "provider": "Anthropic",
      "tier": "A",
      "overall": 87.4,
      "score_type": "Benchquill editorial composite",
      "code": 86.1,
      "reasoning": 87.8,
      "math": 85.6,
      "vision": 84.2,
      "speed_tokens_per_second": 82,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 15,
      "output_price_usd_per_million_tokens": 75,
      "blended_price_usd_per_million_tokens": 60,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "200k",
      "record_note": "",
      "release_month": "2025-05",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 10,
      "rank_basis": "overall_score_descending",
      "slug": "grok-4",
      "name": "Grok 4.20",
      "provider": "xAI",
      "tier": "A",
      "overall": 86.4,
      "score_type": "Benchquill editorial composite",
      "code": 83.2,
      "reasoning": 88.7,
      "math": 87.9,
      "vision": 81.4,
      "speed_tokens_per_second": 144,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 2,
      "output_price_usd_per_million_tokens": 6,
      "blended_price_usd_per_million_tokens": 5,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "2M",
      "record_note": "",
      "release_month": "2026-04",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 11,
      "rank_basis": "overall_score_descending",
      "slug": "claude-sonnet-4-5",
      "name": "Claude Sonnet 4.5",
      "provider": "Anthropic",
      "tier": "A",
      "overall": 86.2,
      "score_type": "Benchquill editorial composite",
      "code": 88.4,
      "reasoning": 86.7,
      "math": 84.1,
      "vision": 83.5,
      "speed_tokens_per_second": 158,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 3,
      "output_price_usd_per_million_tokens": 15,
      "blended_price_usd_per_million_tokens": 12,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "200k",
      "record_note": "",
      "release_month": "2025-09",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 12,
      "rank_basis": "overall_score_descending",
      "slug": "o4-mini",
      "name": "o4-mini",
      "provider": "OpenAI",
      "tier": "A",
      "overall": 85.4,
      "score_type": "Benchquill editorial composite",
      "code": 84.2,
      "reasoning": 89.6,
      "math": 91.8,
      "vision": 76.8,
      "speed_tokens_per_second": 132,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 1.1,
      "output_price_usd_per_million_tokens": 4.4,
      "blended_price_usd_per_million_tokens": 3.58,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "200k",
      "record_note": "",
      "release_month": "2025-04",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 13,
      "rank_basis": "overall_score_descending",
      "slug": "llama-4-maverick",
      "name": "Llama 4 Maverick",
      "provider": "Meta",
      "tier": "A",
      "overall": 84.7,
      "score_type": "Benchquill editorial composite",
      "code": 82.1,
      "reasoning": 85.4,
      "math": 83.8,
      "vision": 87.6,
      "speed_tokens_per_second": 165,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.15,
      "output_price_usd_per_million_tokens": 0.6,
      "blended_price_usd_per_million_tokens": 0.49,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2025-04",
      "license": "open weight",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 14,
      "rank_basis": "overall_score_descending",
      "slug": "deepseek-r2",
      "name": "DeepSeek R2",
      "provider": "DeepSeek",
      "tier": "A",
      "overall": 84.2,
      "score_type": "Benchquill editorial composite",
      "code": 82.1,
      "reasoning": 91.4,
      "math": 92.8,
      "vision": null,
      "speed_tokens_per_second": 88,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.55,
      "output_price_usd_per_million_tokens": 2.19,
      "blended_price_usd_per_million_tokens": 1.78,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2025-08",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 15,
      "rank_basis": "overall_score_descending",
      "slug": "gemini-3-flash-preview",
      "name": "Gemini 3 Flash Preview",
      "provider": "Google",
      "tier": "A",
      "overall": 83.5,
      "score_type": "Benchquill editorial composite",
      "code": 79.8,
      "reasoning": 82.1,
      "math": 84.6,
      "vision": 86.4,
      "speed_tokens_per_second": 412,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.5,
      "output_price_usd_per_million_tokens": 3,
      "blended_price_usd_per_million_tokens": 2.38,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2025-12",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision",
        "Audio"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 16,
      "rank_basis": "overall_score_descending",
      "slug": "gpt-5-mini",
      "name": "GPT-5 mini",
      "provider": "OpenAI",
      "tier": "A",
      "overall": 82.6,
      "score_type": "Benchquill editorial composite",
      "code": 81.4,
      "reasoning": 83.8,
      "math": 84.7,
      "vision": 78.9,
      "speed_tokens_per_second": 286,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.25,
      "output_price_usd_per_million_tokens": 2,
      "blended_price_usd_per_million_tokens": 1.56,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "400k",
      "record_note": "",
      "release_month": "2025-08",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 17,
      "rank_basis": "overall_score_descending",
      "slug": "gpt-4-1",
      "name": "GPT-4.1",
      "provider": "OpenAI",
      "tier": "B",
      "overall": 81.4,
      "score_type": "Benchquill editorial composite",
      "code": 84.6,
      "reasoning": 80.2,
      "math": 79.1,
      "vision": 78.6,
      "speed_tokens_per_second": 142,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 2,
      "output_price_usd_per_million_tokens": 8,
      "blended_price_usd_per_million_tokens": 6.5,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2025-04",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 18,
      "rank_basis": "overall_score_descending",
      "slug": "claude-haiku-4-5",
      "name": "Claude Haiku 4.5",
      "provider": "Anthropic",
      "tier": "B",
      "overall": 80.4,
      "score_type": "Benchquill editorial composite",
      "code": 81.2,
      "reasoning": 79.8,
      "math": 77.6,
      "vision": 78.5,
      "speed_tokens_per_second": 274,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 1,
      "output_price_usd_per_million_tokens": 5,
      "blended_price_usd_per_million_tokens": 4,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "200k",
      "record_note": "",
      "release_month": "2025-10",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 19,
      "rank_basis": "overall_score_descending",
      "slug": "qwen-2-5-max",
      "name": "Qwen 2.5 Max",
      "provider": "Alibaba",
      "tier": "B",
      "overall": 80.4,
      "score_type": "Benchquill editorial composite",
      "code": 79.8,
      "reasoning": 81.2,
      "math": 84.5,
      "vision": null,
      "speed_tokens_per_second": 96,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.85,
      "output_price_usd_per_million_tokens": 3.4,
      "blended_price_usd_per_million_tokens": 2.76,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2025-01",
      "license": "closed",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 20,
      "rank_basis": "overall_score_descending",
      "slug": "deepseek-v3-2",
      "name": "DeepSeek V3.2",
      "provider": "DeepSeek",
      "tier": "B",
      "overall": 79.8,
      "score_type": "Benchquill editorial composite",
      "code": 78.4,
      "reasoning": 80.2,
      "math": 82.6,
      "vision": null,
      "speed_tokens_per_second": 124,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.27,
      "output_price_usd_per_million_tokens": 0.42,
      "blended_price_usd_per_million_tokens": 0.38,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "Historical DeepSeek API record from Dec 2025; use DeepSeek V4-Pro/V4-Flash for current Apr 2026 DeepSeek API pricing comparisons.",
      "release_month": "2025-12",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 21,
      "rank_basis": "overall_score_descending",
      "slug": "gpt-4o",
      "name": "GPT-4o",
      "provider": "OpenAI",
      "tier": "B",
      "overall": 78.6,
      "score_type": "Benchquill editorial composite",
      "code": 76.4,
      "reasoning": 78.2,
      "math": 74.8,
      "vision": 82.4,
      "speed_tokens_per_second": 168,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 2.5,
      "output_price_usd_per_million_tokens": 10,
      "blended_price_usd_per_million_tokens": 8.13,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2024-05",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 22,
      "rank_basis": "overall_score_descending",
      "slug": "gemini-2-0-flash",
      "name": "Gemini 2.0 Flash",
      "provider": "Google",
      "tier": "B",
      "overall": 78.4,
      "score_type": "Benchquill editorial composite",
      "code": 75.6,
      "reasoning": 77.8,
      "math": 78.9,
      "vision": 81.5,
      "speed_tokens_per_second": 446,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.1,
      "output_price_usd_per_million_tokens": 0.4,
      "blended_price_usd_per_million_tokens": 0.33,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2024-12",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 23,
      "rank_basis": "overall_score_descending",
      "slug": "llama-4-scout",
      "name": "Llama 4 Scout",
      "provider": "Meta",
      "tier": "B",
      "overall": 78.2,
      "score_type": "Benchquill editorial composite",
      "code": 76.4,
      "reasoning": 78.9,
      "math": 77.5,
      "vision": 81.2,
      "speed_tokens_per_second": 218,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.08,
      "output_price_usd_per_million_tokens": 0.3,
      "blended_price_usd_per_million_tokens": 0.25,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "10M",
      "record_note": "",
      "release_month": "2025-04",
      "license": "open weight",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 24,
      "rank_basis": "overall_score_descending",
      "slug": "deepseek-v4-flash",
      "name": "DeepSeek V4-Flash",
      "provider": "DeepSeek",
      "tier": "B",
      "overall": 77.8,
      "score_type": "Benchquill editorial composite",
      "code": 75.2,
      "reasoning": 78.4,
      "math": 80.1,
      "vision": null,
      "speed_tokens_per_second": 312,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.14,
      "output_price_usd_per_million_tokens": 0.28,
      "blended_price_usd_per_million_tokens": 0.25,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "1M",
      "record_note": "",
      "release_month": "2026-04",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 25,
      "rank_basis": "overall_score_descending",
      "slug": "mistral-medium-3-1",
      "name": "Mistral Medium 3.1",
      "provider": "Mistral",
      "tier": "B",
      "overall": 77.6,
      "score_type": "Benchquill editorial composite",
      "code": 78.4,
      "reasoning": 76.8,
      "math": 75.2,
      "vision": null,
      "speed_tokens_per_second": 168,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.4,
      "output_price_usd_per_million_tokens": 2,
      "blended_price_usd_per_million_tokens": 1.6,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2025-08",
      "license": "closed",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 26,
      "rank_basis": "overall_score_descending",
      "slug": "grok-4-1-fast",
      "name": "Grok 4.1 Fast",
      "provider": "xAI",
      "tier": "B",
      "overall": 76.8,
      "score_type": "Benchquill editorial composite",
      "code": 74.5,
      "reasoning": 78.2,
      "math": 77.4,
      "vision": 73.1,
      "speed_tokens_per_second": 358,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.2,
      "output_price_usd_per_million_tokens": 0.5,
      "blended_price_usd_per_million_tokens": 0.43,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "2M",
      "record_note": "",
      "release_month": "2025-12",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 27,
      "rank_basis": "overall_score_descending",
      "slug": "qwen-2-5-72b",
      "name": "Qwen 2.5 72B",
      "provider": "Alibaba",
      "tier": "B",
      "overall": 76.8,
      "score_type": "Benchquill editorial composite",
      "code": 75.4,
      "reasoning": 77.6,
      "math": 80.2,
      "vision": null,
      "speed_tokens_per_second": 142,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.35,
      "output_price_usd_per_million_tokens": 1.4,
      "blended_price_usd_per_million_tokens": 1.14,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2024-09",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 28,
      "rank_basis": "overall_score_descending",
      "slug": "command-a",
      "name": "Command A",
      "provider": "Cohere",
      "tier": "B",
      "overall": 76.4,
      "score_type": "Benchquill editorial composite",
      "code": 72.8,
      "reasoning": 78.2,
      "math": 71.5,
      "vision": null,
      "speed_tokens_per_second": 156,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 2.5,
      "output_price_usd_per_million_tokens": 10,
      "blended_price_usd_per_million_tokens": 8.13,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "256k",
      "record_note": "",
      "release_month": "2025-03",
      "license": "closed",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 29,
      "rank_basis": "overall_score_descending",
      "slug": "kimi-k1-5",
      "name": "Kimi K1.5",
      "provider": "Moonshot",
      "tier": "B",
      "overall": 76.4,
      "score_type": "Benchquill editorial composite",
      "code": 73.8,
      "reasoning": 78.6,
      "math": 80.2,
      "vision": 74.1,
      "speed_tokens_per_second": 124,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.8,
      "output_price_usd_per_million_tokens": 2.4,
      "blended_price_usd_per_million_tokens": 2,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "200k",
      "record_note": "",
      "release_month": "2025-01",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 30,
      "rank_basis": "overall_score_descending",
      "slug": "hunyuan-turbo",
      "name": "Hunyuan Turbo",
      "provider": "Tencent",
      "tier": "B",
      "overall": 75.8,
      "score_type": "Benchquill editorial composite",
      "code": 74.2,
      "reasoning": 76.4,
      "math": 75.9,
      "vision": null,
      "speed_tokens_per_second": 156,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.4,
      "output_price_usd_per_million_tokens": 1.2,
      "blended_price_usd_per_million_tokens": 1,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2025-02",
      "license": "closed",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 31,
      "rank_basis": "overall_score_descending",
      "slug": "pixtral-large",
      "name": "Pixtral Large",
      "provider": "Mistral",
      "tier": "B",
      "overall": 75.2,
      "score_type": "Benchquill editorial composite",
      "code": 71.4,
      "reasoning": 74.8,
      "math": 72.6,
      "vision": 82.4,
      "speed_tokens_per_second": 142,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 2,
      "output_price_usd_per_million_tokens": 6,
      "blended_price_usd_per_million_tokens": 5,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2024-11",
      "license": "open weight",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 32,
      "rank_basis": "overall_score_descending",
      "slug": "mistral-large-3",
      "name": "Mistral Large 3",
      "provider": "Mistral",
      "tier": "B",
      "overall": 74.5,
      "score_type": "Benchquill editorial composite",
      "code": 76.2,
      "reasoning": 73.8,
      "math": 71.6,
      "vision": 82.4,
      "speed_tokens_per_second": 124,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.5,
      "output_price_usd_per_million_tokens": 1.5,
      "blended_price_usd_per_million_tokens": 1.25,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "256k",
      "record_note": "",
      "release_month": "2025-12",
      "license": "open weight",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 33,
      "rank_basis": "overall_score_descending",
      "slug": "nova-pro",
      "name": "Nova Pro",
      "provider": "Amazon",
      "tier": "B",
      "overall": 73.8,
      "score_type": "Benchquill editorial composite",
      "code": 72.4,
      "reasoning": 74.6,
      "math": 73.2,
      "vision": 75.8,
      "speed_tokens_per_second": 184,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.8,
      "output_price_usd_per_million_tokens": 3.2,
      "blended_price_usd_per_million_tokens": 2.6,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "300k",
      "record_note": "",
      "release_month": "2024-12",
      "license": "closed",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 34,
      "rank_basis": "overall_score_descending",
      "slug": "hermes-3-405b",
      "name": "Hermes 3 405B",
      "provider": "Nous Research",
      "tier": "B",
      "overall": 72.8,
      "score_type": "Benchquill editorial composite",
      "code": 71.4,
      "reasoning": 73.6,
      "math": 72.9,
      "vision": null,
      "speed_tokens_per_second": 86,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.9,
      "output_price_usd_per_million_tokens": 0.9,
      "blended_price_usd_per_million_tokens": 0.9,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2024-08",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 35,
      "rank_basis": "overall_score_descending",
      "slug": "glm-4-5",
      "name": "GLM-4.5",
      "provider": "Zhipu",
      "tier": "C",
      "overall": 72.6,
      "score_type": "Benchquill editorial composite",
      "code": 71.8,
      "reasoning": 73.4,
      "math": 75.2,
      "vision": null,
      "speed_tokens_per_second": 142,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.2,
      "output_price_usd_per_million_tokens": 0.8,
      "blended_price_usd_per_million_tokens": 0.65,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2024-12",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 36,
      "rank_basis": "overall_score_descending",
      "slug": "phi-4-multimodal",
      "name": "Phi-4-multimodal",
      "provider": "Microsoft",
      "tier": "C",
      "overall": 72.4,
      "score_type": "Benchquill editorial composite",
      "code": 70.6,
      "reasoning": 71.8,
      "math": 73.4,
      "vision": 76.2,
      "speed_tokens_per_second": 286,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.1,
      "output_price_usd_per_million_tokens": 0.3,
      "blended_price_usd_per_million_tokens": 0.25,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2025-02",
      "license": "open weight",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 37,
      "rank_basis": "overall_score_descending",
      "slug": "llama-3-3-70b",
      "name": "Llama 3.3 70B",
      "provider": "Meta",
      "tier": "C",
      "overall": 71.4,
      "score_type": "Benchquill editorial composite",
      "code": 69.8,
      "reasoning": 72.4,
      "math": 70.6,
      "vision": null,
      "speed_tokens_per_second": 178,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.12,
      "output_price_usd_per_million_tokens": 0.3,
      "blended_price_usd_per_million_tokens": 0.26,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2024-12",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 38,
      "rank_basis": "overall_score_descending",
      "slug": "mistral-small-3-1",
      "name": "Mistral Small 3.1",
      "provider": "Mistral",
      "tier": "C",
      "overall": 71.2,
      "score_type": "Benchquill editorial composite",
      "code": 72.4,
      "reasoning": 70.8,
      "math": 69.4,
      "vision": 70.1,
      "speed_tokens_per_second": 234,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.1,
      "output_price_usd_per_million_tokens": 0.3,
      "blended_price_usd_per_million_tokens": 0.25,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2025-03",
      "license": "open weight",
      "modalities": [
        "Text",
        "Vision"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 39,
      "rank_basis": "overall_score_descending",
      "slug": "phi-4",
      "name": "Phi-4",
      "provider": "Microsoft",
      "tier": "C",
      "overall": 70.8,
      "score_type": "Benchquill editorial composite",
      "code": 71.4,
      "reasoning": 70.2,
      "math": 76.8,
      "vision": null,
      "speed_tokens_per_second": 312,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.07,
      "output_price_usd_per_million_tokens": 0.14,
      "blended_price_usd_per_million_tokens": 0.12,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "16k",
      "record_note": "",
      "release_month": "2024-12",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 40,
      "rank_basis": "overall_score_descending",
      "slug": "command-r",
      "name": "Command R+",
      "provider": "Cohere",
      "tier": "C",
      "overall": 70.8,
      "score_type": "Benchquill editorial composite",
      "code": 68.4,
      "reasoning": 71.6,
      "math": 67.8,
      "vision": null,
      "speed_tokens_per_second": 134,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 2.5,
      "output_price_usd_per_million_tokens": 10,
      "blended_price_usd_per_million_tokens": 8.13,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2024-08",
      "license": "closed",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 41,
      "rank_basis": "overall_score_descending",
      "slug": "yi-lightning",
      "name": "Yi-Lightning",
      "provider": "01.AI",
      "tier": "C",
      "overall": 68.4,
      "score_type": "Benchquill editorial composite",
      "code": 66.8,
      "reasoning": 69.2,
      "math": 71.4,
      "vision": null,
      "speed_tokens_per_second": 198,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.14,
      "output_price_usd_per_million_tokens": 0.14,
      "blended_price_usd_per_million_tokens": 0.14,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "16k",
      "record_note": "",
      "release_month": "2024-10",
      "license": "closed",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 42,
      "rank_basis": "overall_score_descending",
      "slug": "aya-expanse-32b",
      "name": "Aya Expanse 32B",
      "provider": "Cohere",
      "tier": "C",
      "overall": 67.8,
      "score_type": "Benchquill editorial composite",
      "code": 65.4,
      "reasoning": 69.1,
      "math": 64.2,
      "vision": null,
      "speed_tokens_per_second": 156,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.5,
      "output_price_usd_per_million_tokens": 1.5,
      "blended_price_usd_per_million_tokens": 1.25,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2024-10",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 43,
      "rank_basis": "overall_score_descending",
      "slug": "gemma-3-27b",
      "name": "Gemma 3 27B",
      "provider": "Google",
      "tier": "C",
      "overall": 67.2,
      "score_type": "Benchquill editorial composite",
      "code": 64.8,
      "reasoning": 68.4,
      "math": 70.1,
      "vision": null,
      "speed_tokens_per_second": 184,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.07,
      "output_price_usd_per_million_tokens": 0.07,
      "blended_price_usd_per_million_tokens": 0.07,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2025-03",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 44,
      "rank_basis": "overall_score_descending",
      "slug": "dbrx",
      "name": "DBRX",
      "provider": "Databricks",
      "tier": "C",
      "overall": 65.4,
      "score_type": "Benchquill editorial composite",
      "code": 64.8,
      "reasoning": 66.1,
      "math": 63.8,
      "vision": null,
      "speed_tokens_per_second": 142,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.75,
      "output_price_usd_per_million_tokens": 2.25,
      "blended_price_usd_per_million_tokens": 1.88,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "32k",
      "record_note": "",
      "release_month": "2024-03",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    },
    {
      "last_verified": "2026-04-29",
      "rank": 45,
      "rank_basis": "overall_score_descending",
      "slug": "llama-3-3-8b",
      "name": "Llama 3.3 8B",
      "provider": "Meta",
      "tier": "D",
      "overall": 58.4,
      "score_type": "Benchquill editorial composite",
      "code": 56.2,
      "reasoning": 59.8,
      "math": 57.4,
      "vision": null,
      "speed_tokens_per_second": 412,
      "speed_type": "estimated throughput; verify with your own prompt and region",
      "input_price_usd_per_million_tokens": 0.06,
      "output_price_usd_per_million_tokens": 0.06,
      "blended_price_usd_per_million_tokens": 0.06,
      "blended_price_formula": "input_price * 0.25 + output_price * 0.75",
      "context_window": "128k",
      "record_note": "",
      "release_month": "2024-12",
      "license": "open weight",
      "modalities": [
        "Text"
      ]
    }
  ]
}