Unverified50% confidenceBenchmarkExact time
7-8B 是lx工具的最低门槛，llama3.1:8b 和 qwen2.5:7b 表现良好，14B级别模型基本能达到云端API效果
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/20/2026
First Seen
Valid until: 10/18/2026
Sources
Related Claims
Unverifiedllama3.1:8b和qwen2.5:7b在lx上表现良好，7-8B是这套工具正常运行的最低门槛85% similarUnverifiedLFM2.5-8B-A1B总参数量为8B，推理时仅激活1.5B参数67% similarUnverified在MiniMax的MMClaw评估中，M2.7在开放式工具调用场景下接近Sonnet 4.6的水平65% similarVerifiedQwen 2.5拥有包括1.5B、7B、72B在内的多个参数版本65% 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 7B65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/573254API
curl https://kongchang.com/api/v1/knowledge/claims/573254MCP
get_claim(id=573254)