Nemotron 3 Super
LLM Provider / Model · Top-tier open-weight agentic coding, 1M-context refactors, GPU-rich self-hosted deployments, NVIDIA ecosystem (NIM/NeMo), governed enterprise environments
At a glance
| Input price | Free (self-hosted) |
| Output price | Free (self-hosted) |
| Price tier | Free |
| Context window | 1M+ |
| Max output | 32K |
| Context tier | 500K+ |
| Speed tier | Fast |
| Latency | local-bound |
| Knowledge cutoff | Jan 2026 |
| Modality | Text-only |
| Model ID | nemotron-3-super-120b-a12b |
| Provider | NVIDIA |
| HumanEval | 89% |
| MMLU | 85% |
| SWE-Bench | 60% |
| Benchmark | SWE-bench Verified: top open-weight |
| Released | Mar 2026 |
| Hosting | Open-weights |
| Capabilities | Tool use, Streaming, Structured output, Hybrid Mamba-Transformer (120B / 12B active), Agentic coding, Function calling, 2.2x throughput vs GPT-OSS-120B, Self-hostable |
What Nemotron 3 Super does
Tool use, Streaming, Structured output, Hybrid Mamba-Transformer (120B / 12B active), Agentic coding, Function calling, 2.2x throughput vs GPT-OSS-120B, Self-hostable
Best for
Top-tier open-weight agentic coding, 1M-context refactors, GPU-rich self-hosted deployments, NVIDIA ecosystem (NIM/NeMo), governed enterprise environments
Works well with
Agent / Orchestration
Conflicts & caveats
- Trae's built-in model menu may not include Nemotron 3 Super — Trae primarily ships with Claude/GPT/Gemini/DeepSeek.
- Google Antigravity is built around Gemini 3 Pro; Claude and recent GPT-5 models also work, but Nemotron 3 Super is not in the supported model list.
- Kiro's spec mode is tuned for Claude models — Nemotron 3 Super may work but spec-driven flows are best with Claude.
- Cursor's built-in model menu may not include Nemotron 3 Super — verify it's available in your Cursor model picker.
- OpenAI Codex (VS Code) is tied to OpenAI's GPT-5.1 Codex family — Nemotron 3 Super is not selectable inside this integration.
- Gemini CLI is locked to Google Gemini models — your LLM is Nemotron 3 Super.
- Privacy conflict: Self-hosted Nemotron 3 Super sends code to cloud @codebase (Cursor). Use local context (Continue indexing, ChromaDB, LanceDB, pgvector, Vespa self-hosted) for true privacy.
- Claude Code only works with Claude models — your LLM is Nemotron 3 Super.
- Devin uses Cognition's proprietary model internally — your selected Nemotron 3 Super will be ignored.
- OpenAI Codex (Cloud) runs GPT-5.1 Codex internally — your selected Nemotron 3 Super will be ignored.
- Jules runs Gemini models internally — your selected Nemotron 3 Super will be ignored.
- The Claude Code SDK is tuned for Claude models — Nemotron 3 Super may work via the model adapter but loses Claude-specific features (thinking, prompt caching, sub-agents).
- Lovable uses a fixed internal model — your selected Nemotron 3 Super will be ignored.
- Replit Agent 3 uses its own model selection internally — your selected Nemotron 3 Super will be ignored.
Build a full stack around Nemotron 3 Super — Flowpicker shows compatibility warnings before you commit.
Open the stack planner →