Runs on ≤16 GB · Hugging Face ↗
Not independently scored by Artificial Analysis
JetBrains' purpose-built coding MoE — 12B total, ~2.5B active, trained on ~10.6T tokens for code completion and infill. The rare coding specialist that actually fits a 16 GB card, and ~2x faster than dense models its size. Custom `MellumForCausalLM` arch.
BigCodeBench-Hard pass@1 28% (34/121), via local GPU (Ollama GGUF, num_ctx 16384) — comparable full-148, serial. 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 12B completion/infill specialist — respectable on hard function-completion for its size and blistering speed (~5s/problem end-to-end on a 4060 Ti via Ollama Q4), though it isn't an agentic coder.
Real GGUF file sizes = weight VRAM. Add the KV cache for your context (≈2 GB at 32K, fp16).
| Quantization | Size |
|---|---|
| Q3_K_M | 6.2 GB |
| IQ4_XS | 6.8 GB |
| Q4_0 | 7.0 GB |
| Q4_K_S | 7.6 GB |
| Q4_1 | 7.7 GB |
| Q4_K_M | 8.2 GB |
| Q4 | 8.3 GB |
| Q5_K_M | 9.3 GB |
| Q2_K | 10.5 GB |
| Q8_0 | 12.9 GB |
| Q6_K | 22.1 GB |
| BF16 | 24.3 GB |
~8 GB at Q4 — fits an 8-16 GB card. Built for fast completion / fill-in-the-middle rather than long agentic chains, so read a single-shot BCB accordingly. Trending in every "best local coding model" roundup this week.