{
 "site": "TopModels",
 "updated": "2026-09-19",
 "sources": [
  {
   "name": "LMArena",
   "boards": [
    "text-overall-style-control",
    "text-coding",
    "text-to-image",
    "text-to-video"
   ],
   "retrieved": "2026-09-19"
  },
  {
   "name": "Artificial Analysis",
   "metric": "Intelligence Index + pricing/speed/context/open-weights",
   "retrieved": "2026-09-19"
  }
 ],
 "categories": {
  "image": {
   "title": "Best image",
   "winner": "GPT-Image 2.5 (Sunburst)",
   "slug": "gpt-image-2-5-sunburst",
   "score": "Arena Score 1421"
  },
  "video": {
   "title": "Best video",
   "winner": "Gemini Omni 1.1 Flash",
   "slug": "gemini-omni-1-1-flash",
   "score": "Arena Score 1515"
  },
  "code": {
   "title": "Best code",
   "winner": "Claude Fable 5",
   "slug": "claude-fable-5",
   "score": "Arena Score 1552"
  },
  "reasoning": {
   "title": "Best reasoning",
   "winner": "Claude Fable 5.1",
   "slug": "claude-fable-5-1",
   "score": "Index 53.4"
  },
  "open": {
   "title": "Best open model",
   "winner": "Kimi K3",
   "slug": "kimi-k3",
   "score": "Arena Score 1485"
  },
  "value": {
   "title": "Best value",
   "winner": "GLM 5.3 Flash",
   "slug": "glm-5-3-flash",
   "score": "176 pts / $1M"
  }
 },
 "value_formula": "Intelligence Index / blended price (3:1), per $1M tokens",
 "value_points_per_million": 176,
 "top10": [
  {
   "rank": 1,
   "name": "Claude Fable 5",
   "org": "Anthropic",
   "url": "/model/claude-fable-5/",
   "arena_score": 1505.7,
   "ci": [
    1500.9,
    1510.4
   ],
   "votes": 30057,
   "price_1m_in_out": [
    10,
    50
   ],
   "intelligence_index": null,
   "open_weights": false,
   "badges": [
    "overall-1",
    "code-1"
   ]
  },
  {
   "rank": 2,
   "name": "Claude Opus 4.6",
   "org": "Anthropic",
   "url": "/model/claude-opus-4-6/",
   "arena_score": 1504.6,
   "ci": [
    1501.0,
    1508.1
   ],
   "votes": 71993,
   "price_1m_in_out": [
    5,
    25
   ],
   "intelligence_index": null,
   "open_weights": false,
   "badges": []
  },
  {
   "rank": 3,
   "name": "Claude Opus 4.7",
   "org": "Anthropic",
   "url": "/model/claude-opus-4-7/",
   "arena_score": 1501.7,
   "ci": [
    1497.9,
    1505.6
   ],
   "votes": 60002,
   "price_1m_in_out": [
    5,
    25
   ],
   "intelligence_index": null,
   "open_weights": false,
   "badges": []
  },
  {
   "rank": 4,
   "name": "Muse Spark 1.2",
   "org": "Meta",
   "url": "/model/muse-spark-1-2/",
   "arena_score": 1499.6,
   "ci": [
    1489.1,
    1510.1
   ],
   "votes": 3227,
   "price_1m_in_out": [
    1.25,
    4.25
   ],
   "intelligence_index": null,
   "open_weights": false,
   "badges": []
  },
  {
   "rank": 5,
   "name": "Claude Fable 5.1",
   "org": "Anthropic",
   "url": "/model/claude-fable-5-1/",
   "arena_score": 1498.5,
   "ci": [
    1490.3,
    1506.6
   ],
   "votes": 5783,
   "price_1m_in_out": [
    10,
    50
   ],
   "intelligence_index": 53.4,
   "open_weights": false,
   "badges": [
    "reasoning-1"
   ]
  },
  {
   "rank": 6,
   "name": "Muse Spark 1.3",
   "org": "Meta",
   "url": "/model/muse-spark-1-3/",
   "arena_score": 1493.1,
   "ci": [
    1484.3,
    1502.0
   ],
   "votes": 4723,
   "price_1m_in_out": [
    null,
    null
   ],
   "intelligence_index": 48.2,
   "open_weights": false,
   "badges": []
  },
  {
   "rank": 7,
   "name": "Gemini 3.8 Flash",
   "org": "Google",
   "url": "/model/gemini-3-8-flash/",
   "arena_score": 1493.0,
   "ci": [
    1484.4,
    1501.6
   ],
   "votes": 5076,
   "price_1m_in_out": [
    0.75,
    3.75
   ],
   "intelligence_index": 41.2,
   "open_weights": false,
   "badges": []
  },
  {
   "rank": 8,
   "name": "Claude Opus 5",
   "org": "Anthropic",
   "url": "/model/claude-opus-5/",
   "arena_score": 1492.9,
   "ci": [
    1488.7,
    1497.1
   ],
   "votes": 42617,
   "price_1m_in_out": [
    5,
    25
   ],
   "intelligence_index": 50.7,
   "open_weights": false,
   "badges": []
  },
  {
   "rank": 9,
   "name": "Muse Spark 1.1",
   "org": "Meta",
   "url": "/model/muse-spark-1-1/",
   "arena_score": 1492.6,
   "ci": [
    1487.7,
    1497.4
   ],
   "votes": 27615,
   "price_1m_in_out": [
    1.25,
    4.25
   ],
   "intelligence_index": null,
   "open_weights": false,
   "badges": []
  },
  {
   "rank": 10,
   "name": "Gemini 3.7 Flash",
   "org": "Google",
   "url": "/model/gemini-3-7-flash/",
   "arena_score": 1489.8,
   "ci": [
    1481.6,
    1498.0
   ],
   "votes": 5640,
   "price_1m_in_out": [
    0.75,
    3.75
   ],
   "intelligence_index": null,
   "open_weights": false,
   "badges": []
  }
 ],
 "models": [
  {
   "slug": "claude-fable-5",
   "display": "Claude Fable 5",
   "oneshot": "The strongest all-rounder in the world right now, by human preference.",
   "best": [
    "#1 overall on LMArena (Arena Score 1506) across 30,000+ blind human votes",
    "#1 on LMArena's coding board (1552) \u2014 the model people prefer for real coding work",
    "1,000,000-token context window"
   ],
   "weak": [
    "Premium pricing: $10 per million input tokens, $50 per million output",
    "Newer variants have fewer votes, so the score band is wider than veterans like Opus 4.6"
   ],
   "for_who": "Demanding creative, analytical and technical work where being right matters more than being cheap.",
   "badges": [
    "overall-1",
    "code-1"
   ],
   "org": "Anthropic",
   "license": "Proprietary",
   "try_url": "https://www.anthropic.com/news/claude-fable-5-mythos-5",
   "arena": {
    "board": "text",
    "rank": 1,
    "score": 1505.7,
    "lo": 1500.9,
    "hi": 1510.4,
    "votes": 30057,
    "price_in": 10,
    "price_out": 50,
    "ctx": 1000000,
    "source_name": "claude-fable-5-high"
   },
   "coding": {
    "rank": 1,
    "score": 1552.4,
    "votes": 8040,
    "source_name": "claude-fable-5-high"
   },
   "aa": null
  },
  {
   "slug": "claude-opus-4-6",
   "display": "Claude Opus 4.6",
   "oneshot": "The proven workhorse: near-top quality with the deepest vote history in the top 3.",
   "best": [
    "Arena Score 1505 \u2014 statistically tied with #1",
    "72,000 human votes behind the score \u2014 the most battle-tested entry near the top",
    "Half the price of Fable 5: $5 / $25 per million tokens"
   ],
   "weak": [
    "Never quite beats its sibling Fable 5 head-to-head",
    "Anthropic's lineup is crowded \u2014 4.6 vs 4.7 vs 5 is a near-tie you must break yourself"
   ],
   "for_who": "Teams that want top-3 quality, a long track record, and a saner bill.",
   "badges": [],
   "org": "Anthropic",
   "license": "Proprietary",
   "try_url": "https://www.anthropic.com/news/claude-opus-4-6",
   "arena": {
    "board": "text",
    "rank": 2,
    "score": 1504.6,
    "lo": 1501.0,
    "hi": 1508.1,
    "votes": 71993,
    "price_in": 5,
    "price_out": 25,
    "ctx": 1000000,
    "source_name": "claude-opus-4-6-high"
   },
   "coding": {
    "rank": 3,
    "score": 1551.3,
    "votes": 18766,
    "source_name": "claude-opus-4-6-high"
   },
   "aa": null
  },
  {
   "slug": "claude-opus-4-7",
   "display": "Claude Opus 4.7",
   "oneshot": "The newest Opus iteration \u2014 statistically a tie with Opus 4.6.",
   "best": [
    "Arena Score 1502, within a few points of #1",
    "60,000 human votes",
    "Same price as 4.6: $5 / $25 per million tokens"
   ],
   "weak": [
    "Hasn't separated itself from Opus 4.6 on the board",
    "The 4.7 name suggests more than the scoreboard shows \u2014 so far"
   ],
   "for_who": "Anthropic shops picking one Opus: try both at the same price and keep the winner.",
   "badges": [],
   "org": "Anthropic",
   "license": "Proprietary",
   "try_url": "https://www.anthropic.com/news/claude-opus-4-7",
   "arena": {
    "board": "text",
    "rank": 3,
    "score": 1501.7,
    "lo": 1497.9,
    "hi": 1505.6,
    "votes": 60002,
    "price_in": 5,
    "price_out": 25,
    "ctx": 1000000,
    "source_name": "claude-opus-4-7-high"
   },
   "coding": {
    "rank": 2,
    "score": 1551.8,
    "votes": 17149,
    "source_name": "claude-opus-4-7-high"
   },
   "aa": null
  },
  {
   "slug": "muse-spark-1-2",
   "display": "Muse Spark 1.2",
   "oneshot": "Frontier-adjacent quality at roughly a tenth of Fable pricing.",
   "best": [
    "Arena Score 1500 \u2014 inside the leading cluster",
    "$1.25 / $4.25 per million tokens \u2014 top-5 quality at a budget price",
    "Meta's highest-ranked model family"
   ],
   "weak": [
    "Only 3,227 votes so far \u2014 a wide confidence band (1489\u20131510) keeps the true rank uncertain",
    "No public speed or context figures recorded yet"
   ],
   "for_who": "Everyday heavy use where quality-per-dollar beats bragging rights.",
   "badges": [],
   "org": "Meta",
   "license": "Proprietary",
   "try_url": "https://developer.meta.com/ai/models/muse-spark/",
   "arena": {
    "board": "text",
    "rank": 4,
    "score": 1499.6,
    "lo": 1489.1,
    "hi": 1510.1,
    "votes": 3227,
    "price_in": 1.25,
    "price_out": 4.25,
    "ctx": null,
    "source_name": "muse-spark-1.2 (xHigh)"
   },
   "coding": {
    "rank": 9,
    "score": 1535.6,
    "votes": 928,
    "source_name": "muse-spark-1.2 (xHigh)"
   },
   "aa": null
  },
  {
   "slug": "claude-fable-5-1",
   "display": "Claude Fable 5.1",
   "oneshot": "The highest measured reasoning score we track.",
   "best": [
    "Artificial Analysis Intelligence Index 53.4 \u2014 the best number we recorded on 2026-09-20",
    "#5 overall on LMArena (1498) with 5,783 votes",
    "1,000,000-token context window"
   ],
   "weak": [
    "Same premium price as Fable 5 ($10 / $50) while ranking slightly lower on overall human preference",
    "Max-effort modes burn reasoning tokens \u2014 watch your bill"
   ],
   "for_who": "Hard analysis, mathematics and long planning chains where pure reasoning depth is the point.",
   "badges": [
    "reasoning-1"
   ],
   "org": "Anthropic",
   "license": "Proprietary",
   "try_url": "https://www.anthropic.com/claude-fable-and-mythos-5-1",
   "arena": {
    "board": "text",
    "rank": 5,
    "score": 1498.5,
    "lo": 1490.3,
    "hi": 1506.6,
    "votes": 5783,
    "price_in": 10,
    "price_out": 50,
    "ctx": 1000000,
    "source_name": "claude-fable-5.1-max"
   },
   "coding": {
    "rank": 32,
    "score": 1518.9,
    "votes": 1184,
    "source_name": "claude-fable-5.1-max"
   },
   "aa": {
    "index": 53.4,
    "price_in": 10,
    "price_out": 50,
    "speed": 69,
    "ctx": 1000000,
    "open": false,
    "license": null,
    "released": "2026-09-01",
    "org": "Anthropic",
    "org_url": "https://www.anthropic.com/",
    "weights_url": null,
    "blended": 20,
    "source_name": "Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)"
   }
  },
  {
   "slug": "muse-spark-1-3",
   "display": "Muse Spark 1.3",
   "oneshot": "Meta's newest Spark generation \u2014 strong, fast, still fresh.",
   "best": [
    "Arena Score 1493 with only 4,723 votes \u2014 a strong start",
    "Measured output speed 223 tokens/second (median) \u2014 quick for its quality class",
    "Intelligence Index 48.2, fourth-best we recorded"
   ],
   "weak": [
    "Newest means least battle-tested \u2014 the confidence band is wide",
    "LMArena hasn't listed its price yet; the $1.25/$4.25 figure on this card comes from Artificial Analysis"
   ],
   "for_who": "Speed-sensitive products that still want top-10 quality.",
   "badges": [],
   "org": "Meta",
   "license": "Proprietary",
   "try_url": "https://developer.meta.com/ai/models/muse-spark/",
   "arena": {
    "board": "text",
    "rank": 8,
    "score": 1493.1,
    "lo": 1484.3,
    "hi": 1502.0,
    "votes": 4723,
    "price_in": null,
    "price_out": null,
    "ctx": null,
    "source_name": "muse-spark-1.3-max"
   },
   "coding": {
    "rank": 8,
    "score": 1536.7,
    "votes": 1346,
    "source_name": "muse-spark-1.3-max"
   },
   "aa": {
    "index": 48.2,
    "price_in": 1.25,
    "price_out": 4.25,
    "speed": 223,
    "ctx": 1000000,
    "open": false,
    "license": null,
    "released": "2026-09-02",
    "org": "Meta",
    "org_url": "https://ai.meta.com/",
    "weights_url": null,
    "blended": 2,
    "source_name": "Muse Spark 1.3 (max)"
   }
  },
  {
   "slug": "gemini-3-8-flash",
   "display": "Gemini 3.8 Flash",
   "oneshot": "The fast daily driver of the top 10.",
   "best": [
    "Arena Score 1493 \u2014 top-10 quality at $0.75 / $3.75 per million tokens",
    "Measured output speed 298 tokens/second (median) \u2014 the fastest model in our top 10",
    "1,048,576-token context window"
   ],
   "weak": [
    "Below the Anthropic top tier on raw quality",
    "Google's Flash naming makes 3.7 vs 3.8 easy to confuse when you buy"
   ],
   "for_who": "High-volume assistants and agents where snappy answers and cost control matter.",
   "badges": [],
   "org": "Google",
   "license": "Proprietary",
   "try_url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
   "arena": {
    "board": "text",
    "rank": 9,
    "score": 1493.0,
    "lo": 1484.4,
    "hi": 1501.6,
    "votes": 5076,
    "price_in": 0.75,
    "price_out": 3.75,
    "ctx": 1048576,
    "source_name": "gemini-3.8-flash-high"
   },
   "coding": {
    "rank": 10,
    "score": 1535.0,
    "votes": 1434,
    "source_name": "gemini-3.8-flash-high"
   },
   "aa": {
    "index": 41.2,
    "price_in": 0.75,
    "price_out": 3.75,
    "speed": 298,
    "ctx": 1000000,
    "open": false,
    "license": null,
    "released": "2026-09-02",
    "org": "Google",
    "org_url": "https://deepmind.google/technologies/gemini/",
    "weights_url": null,
    "blended": 1.5,
    "source_name": "Gemini 3.8 Flash (high)"
   }
  },
  {
   "slug": "claude-opus-5",
   "display": "Claude Opus 5",
   "oneshot": "Anthropic's newest base line \u2014 top-10 quality on a huge vote base.",
   "best": [
    "Arena Score 1493 across 42,617 votes \u2014 a trustworthy score",
    "Intelligence Index 50.7, third-best we recorded",
    "$5 / $25 per million tokens \u2014 half of Fable pricing"
   ],
   "weak": [
    "Sits behind Anthropic's own Fable 5 and Opus 4.6/4.7 on the overall board",
    "Measured output speed 51 tokens/second \u2014 the slowest of our top 10"
   ],
   "for_who": "Teams standardizing on Anthropic's newest generation who want the long-supported base model.",
   "badges": [],
   "org": "Anthropic",
   "license": "Proprietary",
   "try_url": "https://www.anthropic.com/news/claude-opus-5",
   "arena": {
    "board": "text",
    "rank": 10,
    "score": 1492.9,
    "lo": 1488.7,
    "hi": 1497.1,
    "votes": 42617,
    "price_in": 5,
    "price_out": 25,
    "ctx": 1000000,
    "source_name": "claude-opus-5-high"
   },
   "coding": {
    "rank": 12,
    "score": 1533.3,
    "votes": 11403,
    "source_name": "claude-opus-5-high"
   },
   "aa": {
    "index": 50.7,
    "price_in": 5,
    "price_out": 25,
    "speed": 51,
    "ctx": 1000000,
    "open": false,
    "license": null,
    "released": "2026-07-24",
    "org": "Anthropic",
    "org_url": "https://www.anthropic.com/",
    "weights_url": null,
    "blended": 10,
    "source_name": "Claude Opus 5 (Adaptive Reasoning, Max Effort)"
   }
  },
  {
   "slug": "muse-spark-1-1",
   "display": "Muse Spark 1.1",
   "oneshot": "The stability pick: proven scores, cheap, still in the top 10.",
   "best": [
    "Arena Score 1493 across 27,615 votes \u2014 a well-proven rating",
    "$1.25 / $4.25 per million tokens",
    "Two newer Spark generations confirm the family's direction without invalidating it"
   ],
   "weak": [
    "Superseded by Spark 1.2 and 1.3 at similar prices",
    "No public speed or context figures recorded yet"
   ],
   "for_who": "Deployments that value a settled, widely-voted score over the newest label.",
   "badges": [],
   "org": "Meta",
   "license": "Proprietary",
   "try_url": "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/",
   "arena": {
    "board": "text",
    "rank": 11,
    "score": 1492.6,
    "lo": 1487.7,
    "hi": 1497.4,
    "votes": 27615,
    "price_in": 1.25,
    "price_out": 4.25,
    "ctx": null,
    "source_name": "muse-spark-1.1"
   },
   "coding": {
    "rank": 13,
    "score": 1531.2,
    "votes": 7958,
    "source_name": "muse-spark-1.1"
   },
   "aa": null
  },
  {
   "slug": "gemini-3-7-flash",
   "display": "Gemini 3.7 Flash",
   "oneshot": "One honest step behind its sibling 3.8 \u2014 at the same price.",
   "best": [
    "Arena Score 1490 with 5,640 votes",
    "$0.75 / $3.75 per million tokens, 1,048,576-token context",
    "Everything 3.8 does well, at identical pricing"
   ],
   "weak": [
    "Strictly below Gemini 3.8 Flash on the board at the same price \u2014 new projects should start there",
    "Likely to be superseded on Google's fast release cadence"
   ],
   "for_who": "Existing Gemini 3.7 integrations that need a solid, known quantity.",
   "badges": [],
   "org": "Google",
   "license": "Proprietary",
   "try_url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/",
   "arena": {
    "board": "text",
    "rank": 12,
    "score": 1489.8,
    "lo": 1481.6,
    "hi": 1498.0,
    "votes": 5640,
    "price_in": 0.75,
    "price_out": 3.75,
    "ctx": 1048576,
    "source_name": "gemini-3.7-flash-high"
   },
   "coding": {
    "rank": 26,
    "score": 1520.6,
    "votes": 1730,
    "source_name": "gemini-3.7-flash-high"
   },
   "aa": null
  },
  {
   "slug": "gpt-image-2-5-sunburst",
   "display": "GPT-Image 2.5 (Sunburst)",
   "oneshot": "The best image generator in the world, by blind human preference.",
   "best": [
    "#1 on LMArena's text-to-image board (Arena Score 1421)",
    "Beats its own sibling Flare (1399) and the older GPT-Image 2 (1381)",
    "3,149 blind preference votes already"
   ],
   "weak": [
    "The image board is young \u2014 scores will move as votes grow",
    "Per-image pricing not yet published on the board"
   ],
   "for_who": "Anyone generating images where the best-looking result matters more than the cheapest one.",
   "badges": [
    "image-1"
   ],
   "org": "OpenAI",
   "license": "Proprietary",
   "try_url": "https://openai.com/index/introducing-chatgpt-images-2-5/",
   "arena": {
    "board": "text-to-image",
    "rank": 1,
    "score": 1420.7,
    "lo": 1407.6,
    "hi": 1433.8,
    "votes": 3149,
    "price_in": 5,
    "price_out": null,
    "ctx": null,
    "source_name": "gpt-image-2.5-sunburst"
   },
   "coding": null,
   "aa": null
  },
  {
   "slug": "gemini-omni-1-1-flash",
   "display": "Gemini Omni 1.1 Flash",
   "oneshot": "The best video generator in the world, by blind human preference.",
   "best": [
    "#1 on LMArena's text-to-video board (Arena Score 1515)",
    "Leads a Google 1-2 with its sibling Omni Flash (1511)",
    "Ahead of strong open challengers like Wan 3.0 (1494)"
   ],
   "weak": [
    "Young board \u2014 1,777 votes so far, expect movement",
    "Per-second pricing not yet published on the board"
   ],
   "for_who": "Video generation where the result has to look right, not just cheap.",
   "badges": [
    "video-1"
   ],
   "org": "Google",
   "license": "Proprietary",
   "try_url": "https://blog.google/innovation-and-ai/technology/developers-tools/build-with-gemini-omni-1-1-flash/",
   "arena": {
    "board": "text-to-video",
    "rank": 1,
    "score": 1514.6,
    "lo": 1499.1,
    "hi": 1530.1,
    "votes": 1777,
    "price_in": 1.5,
    "price_out": 9,
    "ctx": null,
    "source_name": "gemini-omni-1.1-flash"
   },
   "coding": null,
   "aa": null
  },
  {
   "slug": "kimi-k3",
   "display": "Kimi K3",
   "oneshot": "The strongest open-weights model in the world.",
   "best": [
    "Arena Score 1485 \u2014 ahead of every other non-proprietary model on the overall board",
    "#7 on the coding board (1538) \u2014 open weights that genuinely code",
    "Intelligence Index 43.8, above several closed frontier models; 1,048,576-token context"
   ],
   "weak": [
    "Measured output speed 37 tokens/second \u2014 the slowest card we publish",
    "Pricier than most open rivals: $3 / $15 per million tokens",
    "Custom 'Kimi K3 license' \u2014 check terms before commercial self-hosting"
   ],
   "for_who": "Self-hosters, fine-tuners and anyone who needs the weights, not just the API.",
   "badges": [
    "open-1"
   ],
   "org": "Moonshot",
   "license": "Kimi K3 license",
   "try_url": "https://platform.kimi.ai/docs/guide/kimi-k3-quickstart",
   "arena": {
    "board": "text",
    "rank": 17,
    "score": 1484.8,
    "lo": 1479.5,
    "hi": 1490.0,
    "votes": 20987,
    "price_in": null,
    "price_out": null,
    "ctx": null,
    "source_name": "kimi-k3-max"
   },
   "coding": {
    "rank": 7,
    "score": 1538.4,
    "votes": 5507,
    "source_name": "kimi-k3-max"
   },
   "aa": {
    "index": 43.8,
    "price_in": 3,
    "price_out": 15,
    "speed": 37,
    "ctx": 1048576,
    "open": true,
    "license": "Kimi K3 License",
    "released": "2026-07-16",
    "org": "Kimi",
    "org_url": "https://www.moonshot.cn/",
    "weights_url": "https://huggingface.co/moonshotai/Kimi-K3",
    "blended": 6,
    "source_name": "Kimi K3 (max)"
   }
  },
  {
   "slug": "glm-5-3-flash",
   "display": "GLM 5.3 Flash",
   "oneshot": "The value champion: the most intelligence per dollar we measured.",
   "best": [
    "Intelligence Index 41.9 at $0.15 / $0.50 per million tokens \u2014 176 index points per blended $1M, the best we recorded",
    "MIT license, open weights",
    "1,000,000-token context, ~97 tokens/second median output"
   ],
   "weak": [
    "Not frontier-top on raw quality \u2014 Arena Score 1475 puts it outside the overall top 10",
    "Smaller ecosystem and tooling than the big labs"
   ],
   "for_who": "Startups, agents and pipelines paying per token at scale who refuse to give up real capability.",
   "badges": [
    "value-1"
   ],
   "org": "Z AI",
   "license": null,
   "try_url": "https://huggingface.co/zai-org/GLM-5.3-Flash",
   "arena": null,
   "coding": null,
   "aa": {
    "index": 41.9,
    "price_in": 0.15,
    "price_out": 0.5,
    "speed": 97,
    "ctx": 1000000,
    "open": true,
    "license": "MIT",
    "released": "2026-08-26",
    "org": "Z AI",
    "org_url": "https://z.ai/",
    "weights_url": "https://huggingface.co/zai-org/GLM-5.3-Flash",
    "blended": 0.2375,
    "source_name": "GLM 5.3 Flash"
   }
  }
 ]
}