Runs on ≤12 GB · Hugging Face ↗
Not independently scored by Artificial Analysis
DeepSeek distilled R1 into Qwen2 14B (MIT, Jan 2025). Another baseline, from the first wave of reasoning distillations.
BigCodeBench-Hard pass@1 16% (24/148), via official protocol · llama.cpp Q4_K_M · one request at a time · a shared cluster GPU · mean of 3 runs, range 0.0 pp. 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.
Our own runs on this model. Each number carries how many problems it is over and how many times we ran it — these are not comparable with each other, and none of them is on the same scale as the third-party indices on the leaderboard.
| test | score | over | conditions |
|---|---|---|---|
| BigCodeBench-Hard | 16.2% | n=148 · 3 run(s) | official protocol · llama.cpp Q4_K_M · one request at a time · a shared cluster GPU · mean of 3 runs, range 0. |
Real GGUF file sizes = weight VRAM. Add the KV cache for your context (≈6 GB at 32K, fp16). Read from the GGUF header and checked against what llama.cpp actually allocates.
| Quantization | Size |
|---|---|
| Q2_K | 5.8 GB |
| Q2_K_L | 6.0 GB |
| Q3_K_M | 7.3 GB |
| Q4_K_M | 9.0 GB |
| Q5_K_M | 10.5 GB |
| Q6_K | 12.1 GB |
| Q8_0 | 15.7 GB |
| F16 | 29.5 GB |
16.2% over three runs. It reasons out loud, so it burns tokens: check the budget before putting it in a loop.