Unverified50% confidenceBenchmarkExact time
Running llama.cpp locally on a modern consumer-grade CPU can achieve 10–30 tokens per second.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/6/2026
First Seen
Valid until: 12/5/2026
Sources
Related Entities
Related Claims
UnverifiedA 3B model typically achieves an inference speed of 10–30 tokens/second on a modern CPU via llama.cpp.74% similarUnverifiedllama.cpp通过纯CPU推理和4-bit GGUF量化格式，使7B参数模型在无GPU的普通服务器上达到每秒20-40 tokens的生成速度74% similarVerifiedContinuous Batching improves GPU utilization from the 30–50% of static batching to close to 100%.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/864620API
curl https://kongchang.com/api/v1/knowledge/claims/864620MCP
get_claim(id=864620)