# Nemotron 3 Super

Nemotron 3 Super is an AI model released by NVIDIA on Mar 11 2026. It has 120B parameters, a 1M context window and open source. Tracked results include 54% on BullshitBench v2, 79.2% on GPQA Diamond and 60.5% on SWE-Bench Verified.

## Facts

| Field | Value |
| --- | --- |
| Model | Nemotron 3 Super |
| Developer | NVIDIA |
| Release date | Wednesday, Mar 11 2026 |
| Licensing | Open Source |
| Parameters | 120B |
| Context window | 1M |

## Tracked benchmark scores

| Benchmark | Score | Source | What it measures |
| --- | --- | --- | --- |
| BullshitBench v2 | 54% | [BullshitBench](https://github.com/petergpt/bullshit-benchmark) | Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. |
| SWE-Bench Verified | 60.5% | Lab | Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. |
| GPQA Diamond | 79.2% | Lab | Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. |
| AA Intelligence Index | 36 | Lab | Artificial Analysis composite intelligence index across evals |

## About Nemotron 3 Super

Nemotron 3 Super, released March 11, 2026 at GTC, scaled the Nemotron 3 recipe to 120 billion total parameters with about 12 billion active, and stretched the context window to a native 1 million tokens. The pitch was the accuracy-per-second frontier rather than the accuracy leaderboard alone: NVIDIA reported 478 tokens per second on a single B200 and more than twice the throughput of gpt-oss-120b at matched settings, aimed at long-running agents whose cost is measured in wall-clock and watts.

At launch it posted 79.2% on GPQA Diamond, 60.5% on SWE-Bench Verified using the OpenHands scaffold, and a 36 on the Artificial Analysis Intelligence Index — at the time the strongest showing of any US-developed open-weight model, in a class the Chinese open labs had dominated for a year. Like Nano before it, the release included datasets and training recipes alongside the weights.

## Questions and answers

### When was Nemotron 3 Super released?

Nemotron 3 Super was released by NVIDIA on Wednesday, Mar 11 2026.

### Who made Nemotron 3 Super?

Nemotron 3 Super was built by NVIDIA. The dominant AI chipmaker, publishing the open Nemotron model family. Ships weights, datasets and training recipes to seed demand for its hardware rather than to sell a chatbot.

### What benchmark scores did Nemotron 3 Super get?

Nemotron 3 Super reports 4 tracked benchmark scores — BullshitBench v2: 54%; SWE-Bench Verified: 60.5%; GPQA Diamond: 79.2%; AA Intelligence Index: 36. Tracked scores may come from lab reports or independent benchmarks; source details accompany the benchmark data.

### What is the context window of Nemotron 3 Super?

Nemotron 3 Super has a context window of 1M. That is the maximum amount of input plus output the model can hold in a single request.

### How many parameters does Nemotron 3 Super have?

Nemotron 3 Super is reported at 120B parameters.

### Is Nemotron 3 Super open source?

Yes. Nemotron 3 Super is open source: the model weights are published together with training code and data under an open license, so it can be downloaded, self-hosted, fine-tuned, and in principle reproduced from scratch.

### What came before and after Nemotron 3 Super?

NVIDIA's previous tracked release was Nemotron 3 Nano on Dec 15 2025, 86 days earlier. It was followed by Nemotron 3 Ultra on Jun 4 2026.


---

Canonical page: https://aireleasetracker.com/model/nvidia/nemotron-3-super
Site index: https://aireleasetracker.com/llms.txt
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.