Leaderboard

Popular models leaderboard

Sort and search source-aware performance evidence. Benchmark task rows stay separate from the capability scores, and a value the source did not publish stays unavailable rather than becoming a zero.

Open-weight only

Seven-capability leaderboard

266 source-backed models. Choose a column header to sort; expand a row for its benchmark tasks.

Models scored across seven capability axes

1Claude Opus 5.5Anthropic · Proprietary	86.68	65.3	87.5	85.4	86.1	92.2	Unavailable	97.3
2GPT-6.1 SolOpenAI · Proprietary	84.22	Unavailable	74.1	83.8	90.5	87.3	Unavailable	97.3
3GPT-6 AstraOpenAI · Proprietary	84.04	86.5	76.4	84.3	93.3	87.3	Unavailable	97.3
4Claude Fable 5.1Anthropic · Proprietary	83.22	59.8	91.2	95.4	85.2	Unavailable	Unavailable	96.8
5Gemini 4 ArgonGoogle · Proprietary	82.89	79.6	74	74.9	86.9	Unavailable	Unavailable	92.8
6Claude Sonnet 5.5Anthropic · Proprietary	82.24	78.8	87.5	83.4	81.8	86.7	Unavailable	95.2
7Claude Fable 5Anthropic · Proprietary	81.33	82.9	90.3	85	84.3	66	77.1	96
8Claude Opus 5Anthropic · Proprietary	80.07	85.2	90.4	95.3	81	92.2	Unavailable	91.6
9GPT-6 SolOpenAI · Proprietary	78.6	73.7	77.5	79.9	73.1	86.6	Unavailable	97.3
10GPT-5.6 SolOpenAI · Proprietary	75.29	83.7	64.7	81.6	76.4	89.8	89.1	95.8
11Muse Wonder FullMeta · Proprietary	74.75	Unavailable	69	67.6	82.5	Unavailable	Unavailable	Unavailable
12MiMo-V2.6-ProXiaomi · Open Weight	73.6	77.8	63.3	58.4	82.5	Unavailable	Unavailable	Unavailable
13GPT-5.5 ProOpenAI · Proprietary	73.46	80.4	Unavailable	82	78.9	Unavailable	Unavailable	94
14Muse WonderMeta · Proprietary	71.48	Unavailable	Unavailable	65.3	80.4	Unavailable	Unavailable	Unavailable
15GPT-5.6 TerraOpenAI · Proprietary	70.16	76	64.3	73.7	69.1	84.5	87.1	90.1
16Kimi K3Moonshot AI · Pending	69.95	81.9	74.6	83.9	69.8	92.6	Unavailable	Unavailable
17Claude Opus 4.8Anthropic · Proprietary	69.93	74.1	75.1	87.5	62.8	91.2	75.4	84.4
18Qwen3.8 MaxAlibaba · Open Weight	69.73	82.9	70.6	71.9	87	91.2	94.7	78.9
19Step 5 PreviewStepFun · Open Weight	69.72	78.5	77.6	70.6	82.5	74.6	Unavailable	Unavailable
20GPT-5.5OpenAI · Proprietary	69.18	81.7	55.5	80.8	70.3	75.5	93.3	91.2
21Gemini 3.8 FlashGoogle · Proprietary	69.09	72.6	77.5	80.6	74.3	87.3	Unavailable	69.7
22GPT-5.4 ProOpenAI · Proprietary	68.69	79.3	Unavailable	83.5	78.1	Unavailable	Unavailable	86.5
23Claude Opus 4.7 (Adaptive)Anthropic · Proprietary	68.67	71.8	65.6	82	57.4	53.7	Unavailable	Unavailable
24Ling 3.1 FlashInclusionAI · Open Weight	68.05	82.3	69.1	57.4	82	Unavailable	Unavailable	Unavailable
25GLM-5.3Z.AI · Open WeightSelf-hosted	67.96	74.5	58.6	64.8	79.8	Unavailable	Unavailable	71.5
26Claude Haiku 5.5Anthropic · Proprietary	67.22	Unavailable	64.8	67.3	60.1	Unavailable	Unavailable	79.1
27MiniMax M3.1 FlashMiniMax · Proprietary	67.09	73.7	Unavailable	74.2	79.8	73.5	Unavailable	Unavailable
28Grok 4.6xAI · Proprietary	67.08	77.6	77.3	75.9	61.4	Unavailable	Unavailable	70.1
29Gemini 3.7 FlashGoogle · Proprietary	66.86	64.7	78.6	80.7	81.6	87.2	Unavailable	74.8
30GPT-5.4OpenAI · Proprietary	66.7	80.3	47.1	77.3	64.2	73.1	90.7	82
31GPT-5.2 ProOpenAI · Proprietary	66.68	Unavailable	Unavailable	Unavailable	Unavailable	Unavailable	Unavailable	Unavailable
32GPT-6 LunaOpenAI · Proprietary	66.51	73.3	66.3	68.2	59.2	81.9	Unavailable	83.7
33Muse Spark 1.1Meta · Proprietary	66.08	66.9	60.2	90.8	78.4	81.9	Unavailable	Unavailable
34Muse Spark 1.2Meta · Proprietary	65.88	75.6	62.9	72.2	79.3	Unavailable	Unavailable	Unavailable
35DeepSeek V4.1 Flash 1017DeepSeek · Open Weight	65.66	83.5	62.9	69.7	82.5	77.9	Unavailable	70.1
36Claude Opus 4.7Anthropic · Proprietary	65.4	37.6	69	77.5	31.4	Unavailable	Unavailable	72.9
37Claude Sonnet 5Anthropic · Proprietary	64.89	77.5	62.8	81.2	81.3	82	Unavailable	69.4
38Gemini 3.1 ProGoogle · Proprietary	64.73	69	78.1	77.1	58.2	80.1	Unavailable	65
39MiMo-V2.6-FlashXiaomi · Open Weight	64.57	75.7	64.1	49.9	76	72.2	Unavailable	Unavailable
40GPT-5.6 LunaOpenAI · Proprietary	64.39	70.2	56.4	70.9	58.2	79.6	Unavailable	84

Quality versus cost

Each point is a catalog row with both an overall score and complete token prices. Cost is logarithmic by default; untick for a linear axis.

$ per successful taskLog cost scale

Lowest blended cost

Up to twelve filtered models, ranked on the 75% input / 25% output blend.

73 of these rows are marked self-hosted by the source: open-weight models with no first-party hosted token rate. 11 of them show a hosted rate matched from another provider; the rest report an unavailable price rather than zero.

62 self-hosted models are left out: the source publishes no hosted token rate for them, so they have no comparable per-token cost — not a cost of zero.

Loading page…

Explore models

Top 1-5

Top 6-10

Featured

Capabilities

Benchmarks

Compare

Featured model comparisons

Featured provider comparisons

Articles

Popular models leaderboard

Seven-capability leaderboard

Quality versus cost

Lowest blended cost

Explore models

Top 1-5

Top 6-10

Featured

Capabilities

Benchmarks

Compare

Featured model comparisons

Featured provider comparisons

Articles

Explore models

Top 1-5

Top 6-10

Featured

Capabilities

Benchmarks

Compare

Featured model comparisons

Featured provider comparisons

Articles

Popular models leaderboard

Seven-capability leaderboard

Quality versus cost

Lowest blended cost