Unverified50% confidenceFactExact time
LFM2.5在模拟真实Agent场景的电信测试中提升了74个点，指令跟随IFEval达到91.8，数学推理Math500达到88.8
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
LFM2.5本地部署实测：8B参数碾压GPT-o3s的工具调用能力
bilibiliAGI_Ananas6/1/2026
Related Claims
UnverifiedLFM2.5在多项基准测试中展现出与体量高达其4倍模型相当的竞争力66% similarUnverifiedIEEE 754 双精度浮点数的指数偏移量为1023，实际指数范围为-1022到+102364% similarVerifiedCoT提示将PaLM 540B在GSM8K数学基准测试中的准确率从17.9%提升至58.1%63% similarUnverified盘古2.0 Flash在AIME类数学测试中得分93.3，在HMMT测试中得分91.5，数学能力处于第一梯队62% similarUnverified开发者Ivan测试MTP-LX 0.3.5版本，在数学基准测试中5分30秒内取得了93.3%的正确率62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/49058API
curl https://kongchang.com/api/v1/knowledge/claims/49058MCP
get_claim(id=49058)