⚠️ DEPRECATED — Use prism-coder:27b instead

This model has been replaced by prism-coder-27b.

The 27B achieves 100% BFCL (same as 32B) with:

• 16% smaller (16 GB vs 19 GB)
• Faster inference (28.5 tok/s vs ~22)
• O(n) DeltaNet long context
• Newer Qwen3.5 base (vs Qwen3)

# Use this instead:
ollama pull dcostenco/prism-coder:27b