Models

Best pair · RTX 5090 32GB351 tok/s

Qwen · dense · 2B · 1

Best pair · RTX 5090 32GB173 tok/s

Google / community · MoE · 26B-A4B · 1

Best pair · M5 Max126.5 tok/s

Qwen · dense · unknown · 2

Best pair · 2× DGX Spark94.4 tok/s

community · MoE 35B-A3B · 35B / 3B act · 1

Best pair · M4 Pro85.5 tok/s

Qwen · MoE 35B-A3B · 35B / 3B act · 1

Hugging FaceOpen pairs of this model

Best pair · M4 Pro82.8 tok/s

Qwen · dense · 4B · 1

Best pair · RTX 3060 12GB50 tok/s

empero-ai · MoE distill · 35B / 3B act · 1

Qwen3.8 distilled into Qwen3.6-35B-A3B. Not official Qwen3.8-27B.

Best pair · M4 Pro49.3 tok/s

Qwen · dense · 9B · 1

Best pair · M4 Pro48.3 tok/s

Qwen · dense · 8B · 1

Best pair · RTX 4090 24GB40.7 tok/s

Qwen · dense · 27B · 4

Hugging FaceOpen pairs of this model

Best pair · GTX 1660 SUPER 6GB40 tok/s

ornith-ai · unknown · unknown · 1

Hugging FaceOpen pairs of this model

Best pair · RTX 5060 Ti 16GB14 tok/s

community · MoE post-train · 35B-A3B class · 1

Not a 27B.