Skip to content

AI model leaderboard

A real-world evaluation bringing together many abilities to find models with stronger overall performance.

19 evaluations · 8 organisations · updated 10/01 14:05

Overalltop 30

Ranked on the evidence several public evaluations share
The top 30 models, sortable by column
Model
01GPT-6 AstraOpenAIReleased 2026-09-03 · Evidence-sensitiveCached input ¥6.7Input ¥67.05 · Output ¥335.2388.3
02gpt-6.1-sol—Released — · Evidence-sensitiveCached input —Input — · Output —87.4
03Claude Opus 5.5AnthropicReleased 2026-09-17 · Building upCached input ¥1.34Input ¥26.82 · Output ¥134.0987.3
04Claude Fable 5.1AnthropicReleased 2026-09-01 · Building upCached input ¥1.68Input ¥67.05 · Output ¥335.2385.7
05claude-sonnet-5.5anthropicReleased — · Building upCached input —Input — · Output —84.1
06Claude Fable 5AnthropicReleased 2026-06-09 · Evidence-sensitiveCached input ¥6.7Input ¥67.05 · Output ¥335.2383.2
07Claude Opus 5AnthropicReleased 2026-07-24 · Evidence-sensitiveCached input ¥3.35Input ¥33.52 · Output ¥167.6182.9
08Muse Spark 1.3MetaReleased 2026-09-02 · Evidence-sensitiveCached input —Input ¥8.38 · Output ¥28.4980.4
09Gemini 3.7 FlashGoogleReleased 2026-08-13 · Evidence-sensitiveCached input ¥0.5Input ¥5.03 · Output ¥25.1480.1
10Gemini 3.8 FlashGoogleReleased 2026-09-02 · Evidence-sensitiveCached input ¥0.5Input ¥5.03 · Output ¥25.1479.7
11GPT-5.6 SolOpenAIReleased 2026-07-09 · Evidence-sensitiveCached input ¥2.68Input ¥26.82 · Output ¥134.0979.2
12GPT-5.5OpenAIReleased 2026-04-23 · Evidence-sensitiveCached input ¥3.35Input ¥33.52 · Output ¥201.1472.6
13GPT-6 SolOpenAIReleased 2026-09-22 · Evidence-sensitiveCached input ¥1.34Input ¥13.41 · Output ¥67.0571.3
14Kimi K3Moonshot AIReleased 2026-07-16 · Evidence-sensitiveCached input ¥2Input ¥20 · Output ¥10070.7
15Qwen3.8 MaxAlibabaReleased 2026-07-19 · Evidence-sensitiveCached input —Input ¥12 · Output ¥3670.2
16Claude Opus 4.7AnthropicReleased 2026-04-16 · Evidence-sensitiveCached input ¥3.35Input ¥33.52 · Output ¥167.6169.8
17Muse Spark 1.2MetaReleased 2026-08-05 · Evidence-sensitiveCached input —Input ¥8.38 · Output ¥28.4968.5
18Claude Opus 4.8AnthropicReleased 2026-05-28 · Evidence-sensitiveCached input ¥3.35Input ¥33.52 · Output ¥167.6167.0
19Qwen3.8 Max 0902AlibabaReleased 2026-09-01 · Evidence-sensitiveCached input —Input ¥12 · Output ¥3666.5
20GLM-5.3Z.aiReleased 2026-08-18 · Evidence-sensitiveCached input ¥1.74Input ¥9.39 · Output ¥29.565.6
21Claude Opus 4.6AnthropicReleased 2026-02-05 · Evidence-sensitiveCached input ¥3.35Input ¥33.52 · Output ¥167.6164.8
22Muse Spark 1.1MetaReleased 2026-07-09 · Evidence-sensitiveCached input —Input ¥8.38 · Output ¥28.4961.0
23Gemini 3 Pro PreviewGoogleReleased 2025-11-18 · Evidence-sensitiveCached input —Input — · Output —60.5
24Gemini 3.6 FlashGoogleReleased 2026-07-21 · Evidence-sensitiveCached input ¥0.5Input ¥5.03 · Output ¥25.1459.6
25Gemini 3.5 FlashGoogleReleased 2026-05-19 · Evidence-sensitiveCached input ¥1.01Input ¥10.06 · Output ¥60.3458.2
26Grok 4.6xAIReleased 2026-08-12 · Evidence-sensitiveCached input ¥3.35Input ¥13.41 · Output ¥40.2357.4
27GPT-5.4OpenAIReleased 2026-03-05 · Evidence-sensitiveCached input ¥1.68Input ¥16.76 · Output ¥100.5757.2
28DeepSeek V4.1 FlashDeepSeekReleased 2026-09-10 · Evidence-sensitiveCached input ¥0.0402Input ¥2.01 · Output ¥8.0554.0
29GPT-5.6 TerraOpenAIReleased 2026-07-09 · Evidence-sensitiveCached input ¥1.34Input ¥13.41 · Output ¥80.4553.7
30DeepSeek V4 Pro 0813DeepSeekReleased 2026-08-13 · Evidence-sensitiveCached input ¥0.3Input ¥9 · Output ¥2752.1

Each board shows at most 30 models

The consensus index is not an accuracy rate; tied scores keep the order the shared evidence decides.

How to read this board

Aggregates several public evaluations, and different models have different coverage. Prices play no part in the ranking, a missing evaluation does not count as zero, and the index is not an accuracy rate.

How it is calculated

About prices

API prices come from the providers' own sites, per million tokens. Dollar prices are converted to CNY at the rate of 2026-09-30. The cached price is the input price on a cache hit; cache writes, storage and subscriptions are extra.