Nemotron 3 Ultra
Sold by 6 sellers, timed at 111 tok/s, scored 11.8 on the Intelligence Index, on 8 Design Arena boards, and with results on 1 openly licensed board.
retrieved sha256 · 0bb53a77713f
The same weights served by 3 providers, each at its own list price. OpenRouter's rate is its routed rate across whichever of them it uses, and it need not match any one of them.
What sellers charge
6 sellers · 1.88x from cheapest to dearestOne tick per seller, on a log scale of input price per million from $0.40 at DeepInfra to $0.75 at DeepInfra. 1 seller undercuts the $0.50 listed; 3 are dearer. A faint tick declares no number format. Every seller is named on the endpoint ledger below.
Measured speed
6,305 requests · 6 endpoints · 30-minute windowOpenRouter's own timing of its routed traffic, weighted by requests, over one 30-minute window read . Every route is on the endpoint ledger below.
Every endpoint
6 priced · 6 timed · cheapest first- DeepInfradeepinfra/fp4/flex$0.40$1.76$0.08fp416knot matchednonenonenot matched100%
- DeepInfradeepinfra/fp4 · listed$0.50$2.20$0.10fp416knot matchednonenonenot matched100%
- KTkt · listed$0.50$2.20$0.10not declared236k128 tok/s53.4814ms607 req100%
- BaseTenbaseten/fp4/us$0.60$2.40$0.12fp4183k105 tok/s70264ms3,308 req100%
- Venicevenice/fp8$0.625$3.125$0.1875fp833k116 tok/s55580ms1,829 req96.47%
- DeepInfradeepinfra/fp4/priority$0.75$3.30$0.15fp416knot matchednonenonenot matched96.20%
- DeepInfranot matchednot matchednot matchednot matchedno offer32 tok/s11536ms33 reqno offer
- DeepInfranot matchednot matchednot matchednot matchedno offer120 tok/s71485ms521 reqno offer
- DeepInfranot matchednot matchednot matchednot matchedno offer29 tok/s151.66s7 reqno offer
Each track runs from a route's p90, the rate nine in ten of its requests met or beat, up to its median. Weighted by requests, the figure above reports where the traffic went rather than what one route does: 3 of the 6 timed routes are slower than it, and the middle route runs at 111 tok/s. The widest median-to-p90 gap is DeepInfra's, 2.91x.
DeepInfra publishes 3 offers and had 3 routes timed. The two documents share no route id, so those offers and routes are printed apart rather than paired.
Uptime is OpenRouter's own figure for each seller over the trailing day, read once a day, and the column prints the latest reading, taken . 5 of the 7 sellers with a reading fell under 99% on at least one of the 23 days held, 23 Oct 2026 to 21 Nov 2026: DeepInfra (deepinfra/fp4/flex) at 92.02% on 19 Nov 2026, DeepInfra (deepinfra/fp4/priority) at 93.17% on 17 Nov 2026, Venice at 93.99% on 17 Nov 2026, BaseTen (baseten/fp4) at 97.57% on 12 Nov 2026, and 1 more. Source: OpenRouter's public endpoints API, one reading a day for each seller: 1,211 sellers carried one on the latest of the 23 days held.
Source: OpenRouter's public endpoints API. 1257 sellers across 359 models when the mirror ran, 186 of them served by more than one. List prices, not what anybody paid; a discount is the seller's own. Read . Speed is OpenRouter's own timing over one 30-minute window, read .
What each seller accepts
20 parameters · 6 sellers- frequency_penaltyevery seller
- include_reasoningevery seller
- max_tokensevery seller
- presence_penaltyevery seller
- reasoningevery seller
- reasoning_effortevery seller
- temperatureevery seller
- tool_choiceevery seller
- toolsevery seller
- top_kevery seller
- top_pevery seller
- response_format5 of 6not BaseTen
- stop5 of 6not KT
- repetition_penalty4 of 6not BaseTen and Venice
- structured_outputs4 of 6not BaseTen and Venice
- logit_bias3 of 6not BaseTen, KT, and Venice
- min_p3 of 6not BaseTen, KT, and Venice
- seed3 of 6not BaseTen, KT, and Venice
- logprobs1 of 6only KT
- top_logprobs1 of 6only KT
11 of these 20 are accepted by every one of the 6 sellers that publish a list; the other 9 are declared by some sellers and not others, so what a request sending one of them gets depends on which seller serves it. Each list is the seller's own declaration in OpenRouter's endpoints API, read , and nothing here tested one.
Price position
342 priced modelsBlended $0.925 per million tokens, 3 in to 1 out: cheaper than 156 of the 342 priced models in the catalog. Each bar is one stretch of the catalog on a logarithmic price scale; the mark is this listing.
Measured quality
145 models with an index · 132 on a design arena boardEach mark is one board this model stands on, placed at the share of that board it scores above, and led to its number in the list under the rule. The columns and the ledger below carry the figures.
- 3D scenes · Models1109 Elowin rate 37.9% · source rank 90the board, 113 ranked
- ASCII art · Models1079 Elowin rate 36.3% · source rank 68the board, 67 ranked
- Code · Models1130 Elowin rate 35.3% · source rank 99the board, 120 ranked
- Data visualisation · Models1137 Elowin rate 37.2% · source rank 87the board, 110 ranked
- Game development · Models1126 Elowin rate 35.3% · source rank 89the board, 120 ranked
- SVG · Models1060 Elowin rate 32.0% · source rank 80the board, 83 ranked
- UI components · Models1125 Elowin rate 35.2% · source rank 95the board, 116 ranked
- Websites · Models1128 Elowin rate 34.3% · source rank 104the board, 126 ranked
Position 85 of the 258 models Epoch has placed, which is not the population of either board above, on a scale anchored at Claude 3.5 Sonnet 130 and GPT-5 150. Epoch states that absolute values are meaningless by themselves and that comparisons between models are what the index is for, that values are not linearly related to benchmark accuracy, and that a highly specialised model may score low while being capable in its own domain.
Benchmark results
1 of 6 boards · 23 results| Task | Score |
|---|---|
| AMPS_Hard | 96 |
| code_completion | 69.56 |
| code_generation | 71.83 |
| connections | 86.67 |
| consecutive_events | 18.77 |
| integrals_with_game | 80 |
| javascript | 54.55 |
| logic_with_navigation | 68 |
| math_comp | 96.08 |
| olympiad | 82.58 |
| paraphrase | 77.38 |
| plot_unscrambling | 47.77 |
| python | 45 |
| simplify | 72.28 |
| spatial | 98 |
| story_generation | 69.23 |
| summarize | 74.88 |
| tablejoin | 46.56 |
| tablereformat | 98.04 |
| theory_of_mind | 61.54 |
| typescript | 16.67 |
| typos | 78 |
| zebra_puzzle | 71.25 |
Percent correct on the task the row names, higher is better. All 23 published columns; there is no average to carry, because the site computes its headline in the browser from a weighting this does not reimplement. LiveBench names this run nemotron-3-ultra-550b-a55b, with no effort setting written on it. Read from livebench.ai ↗, which dates its own data . LiveBench, Apache-2.0. Results produced by the LiveBench maintainers on questions refreshed monthly.
1 of the 6 openly licensed boards Vitna mirrors carries a result for Nemotron 3 Ultra, each printed as its publisher published it and none added to another. Epoch AI Benchmarking Hub, MTEB, HELM, Vectara hallucination leaderboard and Aider polyglot have not published one under a name that matches its listing, and names are matched to the catalog exactly and never guessed. A task that is a link opens its column on /capability, with every run scored on it; an effort that is a link opens its run's row there, where the run is ranked on Vitna's scale. Every board, with its licence and what its publisher measures, is on /benchmarks.
Source: OpenRouter (openrouter.ai/models), as of 17 Nov 16:29 UTC. sha256 0bb53a77713f over 853,270 bytes.
Catalog: OpenRouter's public model list.
Intelligence, coding and agentic indexes: Artificial Analysis, as published by OpenRouter.
Elo, win rate and rank: Design Arena, as published by OpenRouter.
Source: each model's own Hugging Face repository (huggingface.co), as of 20 Nov 2026. 162 repositories read, 4 did not answer.
Licence, parameter count, model class and gating: each model's own Hugging Face repository, the one OpenRouter names for the listing.
A licence is the tag the uploader put on the repository, read from nothing else and verified by nobody.
A parameter count is the total the repository's safetensors files declare; for a mixture of experts that is every expert, not the share one token runs through.
Source: the Epoch AI Capabilities Index (epoch.ai/eci), as of 20 Nov 2026. 258 models placed over 2917 benchmark results.
Epoch AI Benchmarking Hub data, by Epoch AI, under CC BY 4.0.
Vitna ran none of these evaluations and adjusted none of these figures.
