Nemotron 3 UltravsMiniMax M3
Across 3 shared benchmarks, MiniMax M3 leads overall: Nemotron 3 Ultra wins 0, MiniMax M3 wins 3, with 0 ties and an average score difference of -22.31.
Nemotron 3 Ultra
NVIDIA · 2026-06-04 · Reasoning model
MiniMax M3
MiniMaxAI · 2026-06-01 · Multimodal model
Nemotron 3 Ultra0 wins(0%)(100%)3 winsMiniMax M3
Benchmark scores
Grouped by capability, sorted by largest gap within each. 3 shared benchmarks.
AI Agent - Information Search
MiniMax M3 1/1| Benchmark | Nemotron 3 Ultra | MiniMax M3 | Diff |
|---|---|---|---|
| BrowseComp | 44.4046 / 54Thinking (With Tools + Internet) | 83.5012 / 54Thinking (With Tools + Internet) | -39.10 |
AI Agent - Tool Usage
MiniMax M3 1/1| Benchmark | Nemotron 3 Ultra | MiniMax M3 | Diff |
|---|---|---|---|
| Terminal-Bench 2.1 | 56.4041 / 44Thinking (With Tools) | 6635 / 44Thinking (With Tools) | -9.60 |
General Knowledge
MiniMax M3 1/1| Benchmark | Nemotron 3 Ultra | MiniMax M3 | Diff |
|---|---|---|---|
| LiveBench | 51.7888 / 115Normal (No Tools) | 70.0240 / 115Deep Thinking (No Tools) | -18.24 |
Specs
| Field | Nemotron 3 Ultra | MiniMax M3 |
|---|---|---|
| Publisher | NVIDIA | MiniMaxAI |
| Release date | 2026-06-04 | 2026-06-01 |
| Model type | Reasoning model | Multimodal model |
| Architecture | MoE | MoE |
| Parameters | 550B | 428B |
| Context length | 1M | 1M |
| Max output | Not available | 512K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Nemotron 3 Ultra | MiniMax M3 |
|---|---|---|
| Text input | Not public | ¥2.1 / 1M tokens |
| Text output | Not public | ¥8.4 / 1M tokens |
| Cache read | Not public | ¥0.42 / 1M tokens |
One or both models have incomplete public pricing.
Summary
- MiniMax M3leads in:AI Agent - Information Search (1/1), AI Agent - Tool Usage (1/1), General Knowledge (1/1)
On average across the 3 shared benchmarks, MiniMax M3 scores 22.31 higher.
Largest single-benchmark gap: BrowseComp — Nemotron 3 Ultra 44.40 vs MiniMax M3 83.50 (-39.10).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.