Model Spend Arena2ND ED.
35B · dense · 2026-08-14

Qwen3.6 35B A3B

Runs on ≤32 GB · Hugging Face ↗

llama.cppvLLMLM Studio

Coding index 41.9 (Artificial Analysis)

A 35B-total MoE with ~3B active — fast decoding for its capability, but all 35B of experts load into VRAM.

First-party test · not the AA coding index

BigCodeBench-Hard pass@1 33% (40/121), via rented RTX PRO 6000 Blackwell 96 GB (vLLM, BF16) — comparable full-148, conc64. The brutal counterpart to HumanEval — where HumanEval saturates near the top, BCB-Hard spreads the field, so this is the number that actually separates coding ability. A verbose reasoner (~4.5K tokens/problem, finish=stop — not truncated). Punches above its AA index.

Sizes on disk

Real GGUF file sizes = weight VRAM.

QuantizationSize
Q2_K12.3 GB
Q3_K_M16.6 GB
IQ4_XS17.7 GB
Q4_K_S20.9 GB
MXFP421.7 GB
Q4_K_M22.1 GB
Q422.4 GB
Q5_K_M26.5 GB
Q8_036.9 GB
Q838.5 GB
Q6_K61.2 GB
BF1669.4 GB

Config tips

~21 GB at Q4; plan for the full size, not the 3B active. MoE offload to CPU lets a 24 GB card run it slower if you are short on VRAM.