Nemotron 3 Ultra
NVIDIA released 2026-06-04 nvidia-open-model
DeepSeek V4 Flash 0731 is both better and cheaper.
It scores +13.5 higher and costs 92% less ($0.105/M against $1.35/M) on this workload — and it does everything this model does.
Hy3 preview is cheaper still (79% less) but drops 512K → 262K context.
The same model via free is 100% cheaper (free/M) — same weights, different latency.
Every price dimension
| Input | $0.600/M |
|---|---|
| Output | $3.60/M |
| Cached input | $0.200/M |
Independent scores
| Intelligence | 38.3 |
|---|---|
| Coding | 49.3 |
| Agentic | 27.5 |
Sources are listed separately rather than averaged. Across the 62 models both have scored they correlate at r = 0.806 — close agreement overall, but four models rank very differently between them, and a blended score would hide exactly those.
Capability
| Context window | 512K |
|---|---|
| Input modes | text |
| Tool use | yes |
| Reasoning | optional |
| Open weights | yes |
Provenance
| Price source | openrouter.ai |
|---|---|
| Fetched | 2026-08-24 |
| Quality data | verified |
| Cross-checked | vendor page |
NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE). Built on a hybrid Transformer-Mamba mixture-of-experts architecture, it...