Unverified50% confidenceBenchmarkExact time
UP主在4G显存电脑上实测MiniCPM-2B生成速度达每秒57个token,首字生成仅0.22秒
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/19/2026
First Seen
Valid until: 12/18/2026
Sources
Related Entities
Related Claims
Unverified在GSM8K上OQ4E用时2450.2秒,8bit高达5471.2秒,耗时差了两倍多77% similarUnverified在配备96GB显存的M2 Max上以1-bit精度运行时,简单任务生成速度稳定在约28-30 tokens/秒75% similarUnverifiedGemma量化模型在高性能笔记本电脑上Token生成速度可达每秒20-50个,纯CPU推理约5-15 tokens/秒,INT4量化下NVIDIA RTX 4060/4070可将推理速度提升3-5倍75% similarUnverified在AMD 7900XT显卡上从Ollama换用LM Studio并重新调参后,token生成速度从每秒5提升到3674% similarUnverifiedvlog/短视频用户认准无裁切4K60p+机内10bit 4:2:2录制,且要看'连续录制无过热限制',很多入门机4K30p以上有30分钟限时或裁切73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/935200API
curl https://kongchang.com/api/v1/knowledge/claims/935200MCP
get_claim(id=935200)