Unverified50% confidenceBenchmarkExact time
Llama 3.2 1B模型在WebGPU下达到55-64 tokens/s，中位回复耗时0.4-0.5秒，比SmolLM2 360M更快。
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/17/2026
First Seen
Valid until: 1/15/2027
Sources
Related Entities
Related Claims
UnverifiedOpenTPU 运行 LFM 2.5 230M 模型可达到约 59 tokens/秒，在 4-bit 量化下提升到 85.8 tokens/秒74% similarUnverifiedThrough pure CPU inference and the 4-bit GGUF quantization format, llama.cpp enables a 7B-parameter model to achieve a generation speed of 20–40 tokens per second on an ordinary server without a GPU.71% similarUnverifiedIn actual benchmark threads for llama.cpp, the 2021 M1 Max achieved a decode write speed of 61 tokens/s, higher than the 2024 M4 Pro's 51 tokens/s.71% similarUnverifiedDeveloper Primitive reported: 91.2 tokens/s with speculative decoding off, reaching 142.6 tokens/s after enabling 3 speculative tokens, approximately a 56% speedup under the same configuration.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1051516API
curl https://kongchang.com/api/v1/knowledge/claims/1051516MCP
get_claim(id=1051516)