服务器渲染的对比结果

Hunyuan-TurboS vs GPT-4o vs Llama3.1-405B Instruct vs Claude 3.5 Sonnet New vs DeepSeek-V3

模型: Hunyuan-TurboS, GPT-4o, Llama3.1-405B Instruct, Claude 3.5 Sonnet New, DeepSeek-V3。基准数量: 9。

Hunyuan-TurboSGPT-4oLlama3.1-405B InstructClaude 3.5 Sonnet NewDeepSeek-V3

模型	MMLU	MMLU Pro	HumanEval	BBH	GPQA Diamond	LiveCodeBench	SimpleQA	MATH	MATH-500
Hunyuan-TurboS	89.5 normal	79 normal	91 normal	92.2 normal	57.5 normal	32 normal	22.8 normal	89.7 normal	-
GPT-4o	88.7 normal	77.9 normal	90 normal	91.7 normal	70.1 normal	35.1 normal	38.2 normal	75.9 normal	75.9 normal
Llama3.1-405B Instruct	88.6 normal	73.4 normal	89 normal	89.2 normal	49 normal	30.2 normal	17.1 normal	73.9 normal	-
Claude 3.5 Sonnet New	88.3 normal	78 normal	93.7 normal	92.6 normal	65 normal	38.7 normal	28.4 normal	78.3 normal	78 normal
DeepSeek-V3	88.5 normal	75.9 normal	89 normal	92.3 normal	59.1 normal	34.6 normal	24.9 normal	87.8 normal	87.8 normal