Prism ML PrismML  · Ternary–Bonsai-27B PrismML Ternary Bonsai 27B · Qwen3.6 27B · ternary 1.71 bpw · via Inference Providers
prism-ml/Ternary-Bonsai-27B-gguf : together

PRISMML · TERNARY BONSAI 27B

Qwen3.6 27B · ternary weights 1.71 bpw · ≈9.4× smaller than FP16 · 95% of full-precision quality

262K context · hybrid-attention · 80.49 avg on 15 thinking-mode benchmarks · MATH-500 99.20 · AIME 2025 90.84 · HumanEval+ 93.90 · BFCL v3 74.41

type a prompt below and press Enter (Shift+Enter for newline). streaming tokens appear here.

bonsai ~ $ 
Enter to send · Shift+Enter for newline · Esc to stop