Home
Storia in 1 fonti
Doubling Qwen3.6-27B on One RTX 3090: ollama llama.cpp + MTP, Lever by Lever (35.7 80.2 tok/s)
A reader on my last post said Ollama was leaving a lot on the table — that a tuned backend with...
Raccontata da
dev.to
Timeline cronologica
- ·
Doubling Qwen3.6-27B on One RTX 3090: ollama llama.cpp + MTP, Lever by Lever (35.7 80.2 tok/s)
A reader on my last post said Ollama was leaving a lot on the table — that a tuned backend with...
- ·
The Prefill Wall: Why MTP's 2 Barely Moves Long-Context Latency (Qwen3.6-27B, RTX 3090)
My MTP post showed multi-token prediction roughly doubling Qwen3.6-27B's generation on a 3090. A...
- ·
Qwen 3.6 35B-A3B for Local AI in 2026: The 24GB VRAM Line That Gets You 120 tok/s
Qwen 3.6 35B-A3B for Local AI in 2026: The 24GB VRAM Line That Gets You 120 tok/s