context.vn

long Bench V2

12 models evaluated

#ModelProviderTypeScore
1Claude Opus 4.5AnthropicClosedclaude-opus-4-5
2Qwen3.5 397BAlibabaOpenqwen3-5-397b
3Qwen3.6 PlusAlibabaClosedqwen3-6-plus
4Nemotron 3 UltraNVIDIA · Open weight61.9%
5Kimi K2.5Moonshot AI · Open weight61%
6GLM-5Z.AI · Open weight60.8%
7Qwen3.5-27BAlibaba · Open weight60.6%
8Qwen3.5-122B-A10BAlibaba · Open weight60.2%
9Agents-A1InternScience · Open weight60.2%
10Qwen3.5-35B-A3BAlibaba · Open weight59%
11DeepSeek V4 Pro BaseDeepSeek · Open weight51.5%
12DeepSeek V4 Flash BaseDeepSeek · Open weight44.7%