Nemotron 3.5 Super: 120B params, 12B active, Mamba-2 hybrid MoE with 1M context
ollama · x · 2026-10-16
Ollama's model page reveals details of NVIDIA's Nemotron 3.5 Super: 120B total parameters with 12B active per token, a hybrid Mamba-2/Transformer MoE backbone, and a C-RADIO vision encoder. It supports vision Q&A on images, charts and documents, tool calling, optional reasoning traces, and a 1M context window (80-90GB locally). Ollama offers one-command integration with Claude Code, OpenCode, Hermes Agent and OpenClaw.
Related event: Nemotron 3.5 Super VL arrives on Ollama with 5x throughput(2 posts)→
More from coding & agent
- Moving agent work from GPUs to CPUs for durable, predictable actions — cramforce · 2026-10-16
- Jev tutorial: 106x cheaper, 180ms yes/no gates replace PASS/FAIL string matching — max_paperclips · 2026-10-16
- Stop giving agents raw API keys: prepaid balances and per-credential caps — Cold_Hall_5384 · 2026-10-16
- Six-month prediction: hiring for generic agent harnesses built from scratch will cool — yi111 · 2026-10-16
- Omarchy ISO wouldn't boot on NVIDIA, so he had Claude patch it — indragie · 2026-10-16
- Dev uses Grok to clean up 3D printer firmware, catches it sneaking in nonsense code — yacineMTB · 2026-10-16