Runs on ≤32 GB · Hugging Face ↗
Coding index 26.8 (Artificial Analysis)
A 30B mixture-of-experts with only 3B active parameters, hybrid Mamba2-Transformer, built by NVIDIA for always-on coding agents (43 programming languages). The small active-parameter count is the point, it keeps generation cheap and lets the model run with expert-offload on a 16 GB card, or fast on one bigger GPU. It is a reasoning model, so it thinks a lot before it answers.
BigCodeBench-Hard pass@1 35% (42/121), via opencode Zen free (nemotron-3.5-lightning-free), reasoning on, streamed; graded over the 121-problem gold-gated subset with the official unittest suite. 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.
Real GGUF file sizes = weight VRAM. Add the KV cache for your context (≈2 GB at 32K, fp16).
| Quantization | Size |
|---|---|
| MXFP4 | 23.2 GB |
| Q4_K_S | 24.5 GB |
| Q4_K_M | 25.3 GB |
| Q4 | 25.5 GB |
| Q5_K_M | 30.2 GB |
| Q8_0 | 35.0 GB |
| Q6_K | 35.0 GB |
| Q8 | 38.6 GB |
| BF16 | 65.9 GB |
Pull the GGUF (unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF) in llama.cpp or Ollama. On a 16 GB card use a Q4 quant with expert-offload (-ot to keep the experts on CPU), or skip local hardware and hit the free opencode Zen endpoint at $0.
Weights ship as a base model plus GGUF k-quants; the aligned version is the one served free on opencode Zen, which is what we measured. Give it a generous output budget, it reasons a lot before the code.