Nemotron 3.5 Bolt 74B-A9B
AvailableNVIDIA Nemotron 3.5 Bolt 74B-A9B (NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB) is a local-agent model released 2026-10-15, derived from Nemotron 3.5 Super VL with 9 active experts per token instead of 22: 74B total / 9B active hybrid Mamba-Transformer MoE, compressed to ~24 GB via native ~2-bit quantization (IQ2-XXS experts, NVFP4 Mamba layers) to fit consumer GPUs. Text, image and video in; up to 256K context. OpenMDW-1.1 open weights; vLLM, SGLang and GGUF (llama.cpp / Ollama / LM Studio). NVIDIA-measured (2-bit Bolt vs BF16 rivals): GPQA Diamond 86.43, Terminal-Bench 2.1 50.98, BrowseComp 66.75 - ahead of Qwen3.6 35B-A3B and Gemma 4 26B-A4B. Artificial Analysis Intelligence Index 21.4. A smaller ~4 GB Bolt E4B was slated for 2026-10-23 (not tracked yet).
Specifications
- License
- Open weights · OpenMDW-1.1
- Weights
- Downloadable
- Architecture
- hybrid
- Parameters
- 74B · 9B active
- Context window
- 262K tokens
- Max output
- —
- Knowledge cutoff
- —
- Price (in / out, $/M)
- —
- Modalities
- TextVisionVideoCode
Benchmarks
No benchmark scores recorded yet. Spotted some? Submit a correction.
Vendor-reported figures are claims until independently verified. See methodology.