Unverified90% confidenceFactTime unknown
Mistral Medium 3.5在HumanEval基准测试中通过率为77.6%
1
Sources
90%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Mistral Vibe完全指南：免费API驱动的终端AI编程助手
bilibili薛定猫AI
Related Entities
Related Claims
Unverified在自研基准PinBash 3测试中，SOL约得78.6%，TERRA约62.9%，LUNA约44.3%65% similarUnverifiedMistral 3.8B 指令模型使用 Q4 量化和 llama.cpp 原生调用时，综合 Score 从 27.3% 提升到 78.3%，区别仅在于是否有 Forge 护栏63% similarUnverified在综合基准测试中，So得55分（约78.6%），Terra得44分（约62.9%），Luna得31分（约44.3%）62% similarUnverified在HumanEval、MBPP、GSM8K等基准测试上，Gemma 7B的表现显著优于同参数量级的LLaMA 2 7B和Mistral 7B62% similarUnverified在Artificial Analysis的Omniscience幻觉指标上，GPT-5.6 Sol Max高达89%，高于Claude Fable的55%和GLM 5.2的28%62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/25125API
curl https://kongchang.com/api/v1/knowledge/claims/25125MCP
get_claim(id=25125)