Runs on ≤16 GB · Hugging Face ↗
Coding index 13.8 (Artificial Analysis)
The dense 14B of the Qwen3 line — a solid mid-small generalist with a reasoning mode.
HumanEval pass@1 63% (19/30), via local GPU (Ollama, Q4_K_M). Measured by us on this hardware, as a rough sanity check — HumanEval is a different, easier, partly-contaminated benchmark than Artificial Analysis’ composite, so it is not comparable to the coding-index column.
BigCodeBench-Hard pass@1 31% (38/121), via local GPU (Ollama, num_ctx 16384, budget 12K) — comparable full-148. 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.
~25 tok/s on RTX 4060 Ti, measured by us.
Real GGUF file sizes = weight VRAM. Add the KV cache for your context (≈5 GB at 32K, fp16).
| Quantization | Size |
|---|---|
| Q3_K_M | 7.3 GB |
| IQ4_XS | 8.1 GB |
| Q4_0 | 8.5 GB |
| Q4_K_S | 8.6 GB |
| Q4_K_M | 9.0 GB |
| Q4 | 9.2 GB |
| Q4_1 | 9.4 GB |
| Q5_K_M | 10.5 GB |
| Q8_0 | 15.7 GB |
| Q2_K | 17.8 GB |
| Q8 | 18.8 GB |
| Q6_K | 25.4 GB |
| BF16 | 29.5 GB |
~9 GB at Q4, fits 16 GB with generous context. Turn thinking off for latency- sensitive agent loops.