Skip to content
All writing

1 min

Optimisations Wednesday, 25% more throughput than Ollama

Smarter & deeper MTP, MTPLX, and less CPU/GPU synchronisation on Qwen.

View MarkdownOpen in ClaudeOpen in ChatGPT

Hello LinkedIn Fam, welcome to Optimisations Wednesday.

Today's headline is a 25% increase in throughput when compared to Ollama.

How did I do this?

  • Smarter & deeper MTP (Qwen's MTP sidecar looks really impressive and I've found acceptance to be quite high)
  • Switched to MTPLX (wasn't significatly faster it's just built specifically for qwen on MLX so provided an easier entry point for my optimisations)
  • Patched MTPLX with:
    • Smarter MTP draft depth (going deeper when confidence is higher)
    • Reduced required synchronisation between the CPU & GPU

The smarter MTP draft depth enabled 4.3 tokens commited per verification run vs 3.2 with standard MTPLX.

Originally posted on LinkedIn