Benchmarks

AI Model Benchmark Leaderboard with comprehensive heatmap, radar chart, and bar chart comparing frontier AI models across 25+ benchmarks.

View Mode

Capability Dimension

Sort Models

Heatmap:

Top (95%+)

Strong (80%+)

Moderate (70%+)

Below avg (50%+)

Weak (<50%)

No data

🧠

5 benchmarks

Model	MMLU-Pro90%	GPQA Diamond94.3%	HLE57.9%	SimpleQA75.6%	SuperGPQA88%
Kimi K2 ThinkingMoonshot	87.1	87.6	-	-	-
GPT-5.4OpenAI	-	92.8	-	-	-
Kimi K2.5Moonshot	87.1	87.6	-	-	-
Claude Opus 4.7Anthropic	88	94.2	54.7	-	-
Step-3.5-FlashStepfun	85.8	-	-	-	-
Gemini 3.1 ProGoogle	85	91.9	45.8	75.6	-
GPT-5.5OpenAI	88	93.6	-	45.8	-
Nemotron Ultra 253BNVIDIA	-	76	-	-	-
Claude Opus 4.8Anthropic	-	93.6	57.9	-	-
Gemini 3 ProGoogle	85	87	45.8	-	-
Llama 4 MaverickMeta	80.5	69.8	-	-	-
Claude Opus 4.6Anthropic	82	91.3	53	57.9	-
DeepSeek V4 Pro (Max)DeepSeek	87.5	90.1	37.7	57.9	-
Qwen 3.5Alibaba	87.8	88.4	28.7	-	-
Mistral LargeMistral	-	43.9	-	-	-
GPT-5.4 ProOpenAI	-	93	43.1	45.3	-
DeepSeek R1DeepSeek	84	71.5	-	-	-
MiMo-V2-FlashXiaomi	84.9	83.7	-	-	-
MiniMax M2.5MiniMax	76.5	85.2	-	-	-
DeepSeek V3.2DeepSeek	85	79.9	-	-	-
Grok 3xAI	-	84.6	-	-	-
GLM-5Zhipu AI	70.4	86	50.4	-	-
Claude Sonnet 4.6Anthropic	79.1	89.9	49	-	-
GPT-oss 120BOpenAI	90	80.9	-	-	-

📐

4 benchmarks

Model	MATH-50099%	AIME 2025100%	HMMT97.7%	FrontierMath52.4%
Kimi K2 ThinkingMoonshot	-	99.1	-	-
GPT-5.4OpenAI	99	99	97	-
Kimi K2.5Moonshot	98	96.1	-	-
Claude Opus 4.7Anthropic	98	99	-	-
Step-3.5-FlashStepfun	99.8	99.8	-	-
Gemini 3.1 ProGoogle	94	100	-	-
GPT-5.5OpenAI	99	99	-	-
Nemotron Ultra 253BNVIDIA	97	72.5	-	-
Gemini 3 ProGoogle	91	96	96	-
Claude Opus 4.6Anthropic	97.6	100	-	-
DeepSeek V4 Pro (Max)DeepSeek	97	95.2	95.2	38.3
Qwen 3.5Alibaba	88	-	-	-
Mistral LargeMistral	93.6	88	-	-
GPT-5.4 ProOpenAI	99	99	97.7	51.7
DeepSeek R1DeepSeek	97.3	87.5	-	-
MiMo-V2-FlashXiaomi	94.1	94.1	-	-
MiniMax M2.5MiniMax	85	86.3	-	-
DeepSeek V3.2DeepSeek	94	89.3	-	-
Grok 3xAI	93.6	93.3	-	-
GLM-5Zhipu AI	88	84	-	-
Claude Sonnet 4.6Anthropic	97.8	52.8	-	-
GPT-oss 120BOpenAI	97.9	97.9	-	-

💻

4 benchmarks

Model	SWE-bench V88.6%	LiveCodeBench91.7%	Terminal-Bench 2.090.2%	Aider Polyglot88%
Kimi K2 ThinkingMoonshot	-	83.1	-	-
GPT-5.4OpenAI	-	-	75.1	-
Kimi K2.5Moonshot	76.8	85	50.8	-
Claude Opus 4.7Anthropic	87.6	80	-	-
Step-3.5-FlashStepfun	74.4	86.4	51	-
Gemini 3.1 ProGoogle	78	81.3	56.2	-
GPT-5.5OpenAI	58.6	84	75.1	-
Nemotron Ultra 253BNVIDIA	-	66.3	-	-
Claude Opus 4.8Anthropic	88.6	-	-	-
Gemini 3 ProGoogle	75	71	-	-
Claude Opus 4.6Anthropic	80.8	76	65.4	-
DeepSeek V4 Pro (Max)DeepSeek	80.6	93.5	-	-
Qwen 3.5Alibaba	76.4	83.6	52.5	-
Mistral LargeMistral	-	82.8	-	-
DeepSeek R1DeepSeek	49.2	65.9	-	-
MiMo-V2-FlashXiaomi	73.4	80.6	38.5	-
MiniMax M2.5MiniMax	80.2	65	42.2	-
DeepSeek V3.2DeepSeek	67.8	74.1	39.6	-
Grok 3xAI	49	79.4	52	-
GLM-5Zhipu AI	77.8	52	56.2	-
Claude Sonnet 4.6Anthropic	79.6	72.4	59.1	-
GPT-oss 120BOpenAI	62.4	60	18.7	-

🤖

5 benchmarks

Model	Tau-Bench80.3%	BrowseComp84%	OSWorld76.3%	BFCL70.9%	WebArena71.6%
GPT-5.4OpenAI	-	82.7	75	-	-
Claude Opus 4.7Anthropic	70.2	-	-	-	-
Gemini 3.1 ProGoogle	85.3	-	-	-	-
GPT-5.5OpenAI	-	82.7	75	-	-
Claude Opus 4.6Anthropic	91.9	84	72.7	70.4	-
DeepSeek V4 Pro (Max)DeepSeek	67.9	83.4	76.2	-	-
Qwen 3.5Alibaba	86.7	78.6	62.2	-	-
MiMo-V2-FlashXiaomi	80.3	58.3	-	-	-
GLM-5Zhipu AI	-	75.9	89.7	70.9	-
Claude Sonnet 4.6Anthropic	91.7	74.7	72.5	-	-

👁️

2 benchmarks

📜

3 benchmarks

Model	RULER82%	LongBench v275%	MRCR v274%
Gemini 3 ProGoogle	82	63.3	-
Claude Opus 4.6Anthropic	56	58	-

📋

2 benchmarks