BenchLM Benchmarks

165 benchmarks · 566 model scores · Data from Jun 2, 2026

All Coding Agentic Reasoning Knowledge Multimodal Math Multilingual Instruction Following External

Coding23 benchmarks

2 models

1DeepSeek V4 Pro BaseDeepSeek76.8%

2DeepSeek V4 Flash BaseDeepSeek69.5%

2 models

1DeepSeek V4 Pro BaseDeepSeek59.2%

2DeepSeek V4 Flash BaseDeepSeek56.8%

4 models

1DeepSeek V4 Pro (Max)DeepSeek3206.0

2DeepSeek V4 Flash (Max)DeepSeek3052.0

3DeepSeek V4 Pro (High)DeepSeek2919.0

4DeepSeek V4 Flash (High)DeepSeek2816.0

49 models

1Claude Mythos PreviewAnthropic93.9%

2Claude Opus 4.8Anthropic88.6%

3Claude Opus 4.7 (Adaptive)Anthropic87.6%

4GPT-5.3 CodexOpenAI85%

5Claude Opus 4.5Anthropic80.9%

13 models

1Claude Opus 4.6Anthropic65.3%

2GLM-5Z.AI62.8%

3GLM-5.1Z.AI62.7%

4DeepSeek V3.2DeepSeek60.9%

5Claude Sonnet 4.6Anthropic60.7%

live Code Bench

14 models

1DeepSeek V4 Pro (Max)DeepSeek93.5%

2Qwen3.7 MaxAlibaba91.6%

3DeepSeek V4 Flash (Max)DeepSeek91.6%

4DeepSeek V4 Pro (High)DeepSeek89.8%

5Kimi K2.6Moonshot AI89.6%

live Code Bench V6

8 models

1Kimi K2.6Moonshot AI89.6%

2Qwen3.6 PlusAlibaba87.1%

3Kimi K2.5Moonshot AI85.0%

4Claude Opus 4.5Anthropic84.8%

5Qwen3.5 397BAlibaba83.6%

live Code Bench Pro

6 models

1GPT-5.4OpenAI87.5%

2Gemini 3.1 ProGoogle82.9%

3Muse SparkMeta80.0%

4Grok 4.20xAI74.2%

5Claude Opus 4.6Anthropic70.7%

35 models

1Claude Mythos PreviewAnthropic77.8%

2Claude Opus 4.8Anthropic69.2%

3Claude Opus 4.7 (Adaptive)Anthropic64.3%

4Qwen3.7 MaxAlibaba60.6%

5MiniMax M3MiniMax59%

swe Multilingual

20 models

1Claude Opus 4.8Anthropic84.4%

2Composer 2.5Cursor79.8%

3Qwen3.7 MaxAlibaba78.3%

4Claude Opus 4.5Anthropic77.5%

5Kimi K2.6Moonshot AI76.7%

1 models

1Claude Opus 4.8Anthropic38.4%

7 models

1Claude Opus 4.7 (Adaptive)Anthropic64.8%

2Composer 2.5Cursor63.2%

3GPT-5.5OpenAI59.2%

4Composer 2Cursor52.2%

5Gemini 3.5 FlashGoogle49.8%

multi Swe Bench

1 models

1MiniMax M2.7MiniMax52.7%

1 models

1MiniMax M2.7MiniMax55.6%

8 models

5Claude Opus 4.6claude-opus-4-6-programbenchClosed

7Claude Sonnet 4.6claude-sonnet-4-6-programbenchClosed

8GPT-5.4gpt-5-4-programbenchClosed

9Gemini 3.1 Progemini-3-1-pro-programbenchClosed

10Gemini 3 Flashgemini-3-flash-programbenchClosed

8 models

1Qwen3.7 MaxAlibaba47.2%

2Claude Opus 4.5Anthropic43.2%

3Qwen 3.6 Max (preview)Alibaba42.9%

4GLM-5.1Z.AI42.7%

5MiniMax M3MiniMax42.1%

react Native Evals

16 models

1Composer 2Cursor96.1%

2Composer 2 FastCursor94.9%

3GPT-5.4OpenAI85.3%

4GPT-5.5OpenAI84.7%

5Claude Opus 4.6Anthropic84.1%

swe Verified Arcee

5 models

1Claude Opus 4.6Anthropic75.6%

2MiniMax M2.7MiniMax75.4%

3GLM-5Z.AI72.8%

4Kimi K2.5Moonshot AI70.8%

5Trinity-Large-ThinkingArcee AI63.2%

1 models

1Interfaze BetaInterfaze52.9%

9 models

1Qwen3.7 MaxAlibaba53.5%

2Gemini 3.5 FlashGoogle53.1%

3Kimi K2.6Moonshot AI52.2%

4Kimi K2.5Moonshot AI48.7%

5Grok 4.3xAI47.3%

aa Coding Index

119 models

1GPT-5.5OpenAI59.1%

2GPT-5.4OpenAI57.3%

3Gemini 3.1 ProGoogle55.5%

4GPT-5.3 CodexOpenAI53.1%

5Claude Opus 4.7Anthropic53.1%

122 models

1Gemini 3.1 ProGoogle58.9%

2GPT-5.4OpenAI56.6%

3GPT-5.5OpenAI56.1%

4Gemini 3 ProGoogle56.1%

5GPT-5.2-CodexOpenAI54.6%

terminal Bench Hard

115 models

1GPT-5.5OpenAI60.6%

2GPT-5.4OpenAI57.6%

3Claude Opus 4.7Anthropic54.5%

4Gemini 3.1 ProGoogle53.8%

5GPT-5.3 CodexOpenAI53.0%