Nemotron 3 Ultra 550B A55B
#24 of 31 text models on overall (provisional). $1.86 per 1,000 support replies at list price.
Price checked 25 Oct 2026 via models.dev- Released
- 4 Jun 2026
- Context window
- 262K tokens
- Longest reply
- 183K tokens
- Open weights
- Yes
- Reasoning mode
- Yes
- Tool use
- Yes
- Reads
- text
- Writes
- text
- Knowledge cutoff
- 2025-09
Overall34
Writing43
Support38
Coding51
Value30
Provisional, 0 to 100 among the text models we track, from Arena and Epoch AI. How we score.Prices and where to run it
| Provider | Region | Input per 1M | Output per 1M | Cached input | Data retention | Checked |
|---|---|---|---|---|---|---|
| openrouter | global | $0.60 | $2.40 | $0.12 | Not checked | 25 Oct 2026, models.dev |
Price history and a CSV download arrive with the open price dataset.
Benchmarks behind the scores
| Measure | Figure | Run date | Source |
|---|---|---|---|
| Epoch Capabilities Index | 146 | 25 Oct 2026 | Epoch AI |
| Arena text, coding | 1467.3 | 22 Oct 2026 | Arena |
| Arena text, creative writing | 1398.3 | 22 Oct 2026 | Arena |
| Arena text, instruction following | 1418.1 | 22 Oct 2026 | Arena |
| Arena text, multi-turn | 1411.4 | 22 Oct 2026 | Arena |
| Arena text, overall | 1444.7 | 22 Oct 2026 | Arena |
Call it through OpenRouter
curl https://openrouter.ai/api/v1/chat/completions \
-H "Authorization: Bearer $OPENROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "nvidia/nemotron-3-ultra-550b-a55b", "messages": [{"role": "user", "content": "Hello"}]}'