The full scored language board, including the specialised products listed under Other. Official publisher releases only; quantizations and repacks stay discoverable, unranked.
1 Increased by 1 Current-model rank | Google | 87.95 public benchmark observations across 5 boards; Index basis Measured; 68% support; score interval 83.4–92.4. | 1M 944K route cap 1,048,576 token context; 943,718 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,523 LMArena Highest supported configuration: High; source rank 2 | $10.00 |
2 Decreased by 1 Current-model rank | Anthropic | 86.84 public benchmark observations across 4 boards; Index basis Measured; 66% support; score interval 82.1–91.5. | 1M 128K route cap 1,000,000 token context; 128,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,505 LMArena Highest supported configuration: High; source rank 4 | $20.00 |
3 Increased by 1 Current-model rank | Meta | 86.43 public benchmark observations across 3 boards; Index basis Measured; 63% support; score interval 81.4–91.4. | 1M 944K route cap 1,048,576 token context; 943,718 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,513 LMArena Highest supported configuration: Max; source rank 3 | $20.00 |
4 Decreased by 1 Current-model rank | Anthropic | 86.04 public benchmark observations across 4 boards; Index basis Measured; 67% support; score interval 81.4–90.6. | 1M 128K route cap 1,000,000 token context; 128,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,504 LMArena Highest supported configuration: Max; source rank 6 | $50.00 |
5Current-model rank | Meta | 83.93 public benchmark observations across 3 boards; Index basis Measured; 66% support; score interval 79.1–88.7. | 1M 944K route cap 1,048,576 token context; 943,718 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,495 LMArena Highest supported configuration: Max; source rank 11 | $4.25 |
6Current-model rank | Google | 83.65 public benchmark observations across 5 boards; Index basis Measured; 69% support; score interval 79.1–88.1. | 1M 66K route cap 1,048,576 token context; 65,536 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,498 LMArena Highest supported configuration: High; source rank 9 | $4.13 |
7Current-model rank | OpenAI | 82.4higher pro tier of GPT-6.1 Sol (81.7) (96%); shrunk 1.3 toward the measured corpus; Index basis Estimated; 20% support; score interval 76.2–88.7. | 1M 128K route cap 1,050,000 token context; 128,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | — No Arena rating | $10.00 |
8Current-model rank | OpenAI | 81.75 public benchmark observations across 5 boards; Index basis Measured; 66% support; score interval 77.0–86.4. | 1M 128K route cap 1,050,000 token context; 128,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,483 LMArena Highest supported configuration: Max; source rank 24 | $10.00 |
9Current-model rank | OpenAI | 81.5higher pro tier of GPT-6 Astra (80.7) (96%); shrunk 1.2 toward the measured corpus; Index basis Estimated; 20% support; score interval 75.2–87.7. | 1M 128K route cap 1,050,000 token context; 128,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | — No Arena rating | $50.00 |
10 Increased by 2 Current-model rank | Z.ai | 80.9higher prime tier of GLM 5.3 (79.6) (96%); shrunk 0.7 toward the measured corpus; Index basis Estimated; 20% support; score interval 74.7–87.2. | 1M 131K route cap 1,000,000 token context; 131,072 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | — No Arena rating | $8.80 |
11Current-model rank | Moonshot AILocal model | 80.84 public benchmark observations across 4 boards; Index basis Measured; 68% support; score interval 76.2–85.4. | 1M 944K route cap 1,048,576 token context; 943,718 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,488 LMArena Highest supported configuration: Max; source rank 19 | $16.00 |
12 Increased by 1 Current-model rank | OpenAI | 80.75 public benchmark observations across 5 boards; Index basis Measured; 68% support; score interval 76.1–85.3. | 1M 128K route cap 1,050,000 token context; 128,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,475 LMArena Highest supported configuration: Max; source rank 36 | $50.00 |
13 Increased by 1 Current-model rank | OpenAI | 80.7higher pro tier of GPT-5.5 (80.1) (93%); shrunk 1.2 toward the measured corpus; Index basis Measured; 2% support; score interval 74.5–86.8. | 1M 128K route cap 1,050,000 token context; 128,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | — No Arena rating | $180 |
14 Decreased by 4 Current-model rank | Google | 80.63 public benchmark observations across 3 boards; Index basis Measured; 66% support; score interval 75.9–85.3. | 1M 66K route cap 1,048,576 token context; 65,536 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,487 LMArena Highest supported configuration: Default; source rank 20 | $12.00 |
15 Increased by 2 Current-model rank | Meta | 80.2publisher-corpus-prior over 210 held-out anchors (18%); launch card against 5 resolved peers on 11 rows (82%); exceeds every named peer on 3 of 11 rows; optimism haircut 0 from cross-lab-probe; shrunk 0.7 up toward the measured corpus; Index basis Estimated; 40% support; score interval 67.3–93.0. | 1M 944K route cap 1,048,576 token context; 943,718 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | — No Arena rating | $4.25 |
16 Decreased by 1 Current-model rank | 80.2higher prime tier of Qwen3.8 Max (0803) (79.5) (96%); shrunk 1.3 toward the measured corpus; Index basis Estimated; 20% support; score interval 74.0–86.5. | 1M 131K route cap 1,000,000 token context; 131,072 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | — No Arena rating | $12.00 | |
17 Decreased by 1 Current-model rank | OpenAI | 80.14 public benchmark observations across 4 boards; Index basis Measured; 67% support; score interval 75.4–84.8. | 1M 128K route cap 1,050,000 token context; 128,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,481 LMArena Highest supported configuration: High; source rank 28 | $30.00 |
18 Increased by 2 Current-model rank | 79.63 public benchmark observations across 3 boards; Index basis Measured; 66% support; score interval 74.9–84.3. | 1M 944K route cap 1,048,576 token context; 943,718 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,482 LMArena Highest supported configuration: Max; source rank 26 | $1.40 | |
19Current-model rank | XiaomiLocal model | 79.63 public benchmark observations across 3 boards; Index basis Measured; 54% support; score interval 73.9–85.3. | 1M 131K route cap 1,050,000 token context; 131,072 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,479 LMArena Highest supported configuration: Default; source rank 29 | $0.87 |
20 Decreased by 2 Current-model rank | Anthropic | 79.25 public benchmark observations across 5 boards; Index basis Measured; 67% support; score interval 74.5–83.9. | 1M 128K route cap 1,000,000 token context; 128,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,473 LMArena Highest supported configuration: X-High; source rank 45 | $10.00 |
21Current-model rank | MeituanLocal model | 79.0global-corpus-prior over 210 held-out anchors (6%); launch card against 4 resolved peers on 6 rows (94%); no cross-lab optimism probe available; shrunk 1.6 toward the measured corpus; Index basis Estimated; 32% support; score interval 71.6–86.5. | 1M 262K route cap 1,048,756 token context; 262,144 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | — No Arena rating | $1.20 |
22Current-model rank | 78.14 public benchmark observations across 4 boards; Index basis Measured; 68% support; score interval 73.5–82.7. | 1M 944K route cap 1,048,576 token context; 943,717 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,476 LMArena Highest supported configuration: Default; source rank 33 | $0.50 | |
23Current-model rank | 77.82 public benchmark observations across 2 boards; Index basis Measured; 62% support; score interval 72.8–82.8. | 2M 1.8M route cap 2,000,000 token context; 1,800,000 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | 1,471 LMArena Highest supported configuration: Beta · 0309; source rank 48 | $2.50 | |
24Current-model rank | ByteDance Seed | 77.5global-corpus-prior over 210 held-out anchors (10%); launch card against 2 resolved peers on 3 rows (90%); exceeds every named peer on 1 of 3 rows; no cross-lab optimism probe available; shrunk 2.6 toward the measured corpus; Index basis Estimated; 23% support; score interval 67.7–87.2. | 262K 236K route cap 262,144 token context; 235,929 token catalogue-declared route output cap (OpenRouter; provider/prompt dependent) | — No Arena rating | $2.50 |
25 Increased by 1 Current-model rank | Nex AGILocal model | 77.0global-corpus-prior over 210 held-out anchors (6%); launch card against 6 resolved peers on 5 rows (94%); optimism haircut 0 from cross-lab-probe; shrunk 1.5 toward the measured corpus; Index basis Estimated; 27% support; score interval 69.6–84.5. | — — route cap Unknown context; Catalogue-declared route output cap unknown (OpenRouter; provider/prompt dependent) | — No Arena rating | Self-hosted: weights only · no listed API price |
The first ten positions on the ModelCap Index as of 7 November 2026, with the listed output price and context window beside each. The full board above is searchable and re-orders itself on every refresh.
Public evidence only. Measured positions come from named public benchmark boards — Arena text and coding, LMArena Agent, ARC-AGI, SWE-bench, BFCL and others — matched to an exact model identity. A model those boards have not measured yet is placed from other public evidence instead: the model it is built on, the publisher-reported results on its model card against measured peers, a measured predecessor or tier sibling, or a calibrated prior. On this snapshot, 113 of the 352 ranked positions are measured and 239 are modeled. Downloads, hype and provider counts never move a rank.
Uncertainty is published. Each score carries an interval and an evidence label (Measured, Measured · blended, or Modeled with the channel that placed it), so a one-point gap between neighbours reads as what it is. The methodology and the public dataset let anyone re-derive the board.
Rank, price and availability together. Listed API prices, context windows, weight access and every provider’s endpoint are read from the same validated snapshot as the rank; the changes feed shows the rank moves it can verify against a comparable baseline.
Gemini 4 Argon from Google holds ModelCap Index position #1 as of 7 November 2026 with a score of 87.9, ahead of Claude Opus 5.5 (#2) and Muse Wonder (#3). The Index combines measured results from public benchmark boards with labelled modeled placements and published uncertainty. Small gaps need to be read alongside the intervals and evidence labels; this answer describes the published snapshot shown on the page.
Every current, canonical language model with matched public benchmark evidence receives a ModelCap Index score (0–100) with an interval. Measured evidence comes from named public boards — Arena text and coding, LMArena Agent, ARC-AGI, SWE-bench, BFCL and others — and each result counts by its board's share of its family and by its confidence, which falls as the evidence ages. A model not yet measured on those boards can hold a modeled position instead: carried over from the model it is built on, placed from the publisher-reported results on its model card against measured peers, opened from a measured predecessor or tier sibling, or taken from a calibrated prior. A modeled position is labelled with the channel that placed it and shown with its confidence and interval; it is never presented as a measurement. On this snapshot, 113 of the 352 ranked positions are measured (9 of them blended with modeled terms) and 239 are modeled: 119 carried over from the model they are built on, 101 placed from publisher-reported model-card results against measured peers, 9 opened from a measured predecessor, 6 opened from the measured model they are a tier of and 4 taken from the calibrated prior. The methodology page publishes the weighting, the placement and identity rules and the change log. Market signals (downloads, provider counts) never move a rank.
Gemini 4 Argon from Google leads ModelCap's best-for-coding list as of 7 November 2026 (ModelCap Index #1). The list ranks models measured on at least two admitted coding boards (Arena Coding, SWE-bench bash-only, Terminal-Bench 2.1 with Terminus 2) by the Index's coding component, then lists models measured on a single board by that board's score, with the figures beside each row.
Among ranked models with a positive listed price, Ling 3.0 Tiny is the cheapest by blended token price as of 7 November 2026 ($0.022 per 1M total tokens at a 3:1 input-to-output ratio), at ModelCap Index #139. The cheapest-LLMs list orders priced models by 0.75 × input + 0.25 × output rate; your workload may have a different cheapest option. Separately, 11 ranked models currently have a free tier on at least one provider.
Muse Spark Open from Meta is the highest-ranked model whose weights are open or gated as of 7 November 2026, at ModelCap Index #15; its weights download after accepting the publisher's terms. The open-weight list orders every such model by Index position with its licence note and listed price: open means an OSI or free-culture licence with no access gate, gated means a download after accepting the publisher's terms. Ungated weights under a custom or use-restricted licence are labelled Restricted license and left off the list, which is why Kimi K3 (#11) does not lead it although its weights are downloadable. ModelCap labels weight availability, not whether the full training stack is open source.
Grok 4.20 Multi-Agent lists the largest published context window among ranked models as of 7 November 2026: 2M tokens. The context-window list orders every ranked model by published limit; individual providers can serve less than the maximum.
LMArena's text board is one of ModelCap's inputs - Gemini 4 Argon leads among ModelCap's eligible models with matched results at 1523 as of 7 November 2026 - but the ModelCap Index combines several public boards with published uncertainty and adds listed prices, context windows and provider observations, so one page answers rank, cost and availability together. Each board also has its own leaderboard page here.
AWS runs scheduled discovery checks hourly after the previous cycle finishes, with routine full ranking refreshes when the accepted dataset is six hours old and earlier refreshes for new releases or newly available launch evidence. Only validated snapshots publish; this page uses the snapshot published 7 November 2026. Pages read published snapshots on request through their caches, and the live board checks for updates. Discovery, ranking computation, publication and page refresh are separate steps; there is no guaranteed one-minute update time for every page.
2219 canonical models across language, image, audio and video catalogues, 352 of them with a public ModelCap Index position, served by 84 inference providers observed through OpenRouter as of 7 November 2026. The models directory lists every one, A–Z by publisher.
