The letter

One letter a week, starting soon.

The week's signal, what shipped, what is worth running tonight, and the editor's note. No tracking, no ads, nothing else.

We keep your address, your language and the date you joined, nothing else. Every letter has a one-click unsubscribe link that deletes the record.

← Accept All   Archive
Qiita (LLM)Japan

LM Studio 10 tok/s → 同梱のllama-serverを直接起動したら26 tok/s。--n-cpu-moe 32で30 tok/sまで伸びた(VRAM 8GB)

11.11

前回の記事で、RTX 3060 Ti(VRAM 8GB)で Qwen3.6-35B-A3B を動かすと、同じ GGUF でも LM Studio が 6.2 tok/s、Ollama が 19.1 tok/s になったと書きました。原因は分からないままで、「llama.c...

Models & releasesOpen weights
Read at Qiita (LLM) ↗

Related

More from Qiita (LLM) on Accept All.