Kapitsa:Apple Silicon 推理引擎,同机 69 到 161 tok/s
Stars 373许可 Apache-2.0吞吐 69 → 161 tok/s (claimed)
Why it matters
面向 Apple Silicon 的本地推理引擎,标题给出同机吞吐对比数字。
Summary
Blazing fast Apple Silicon optimized inference engine built for Qwen & Nemotron
Key points
- 面向 Qwen 与 Nemotron,兼容 OpenAI / Anthropic API(仓库描述)
- 标题称同机吞吐约 69 → 161 tok/s
Comments
Comments coming soon.