🔍 Read the full analysis: What To Know About Nemotron 3.5 Super VL’s 25-Cent Tasks on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
NVIDIA released Nemotron 3.5 Super VL on Oct. 15, 2026. Artificial Analysis rates it at 26.5 on its Intelligence Index and estimates a benchmark task costs $0.25, while provider speed varies widely. The model’s case rests on speed, commercial-use licensing and self-hosting options; the supplied figures do not establish it as the strongest model for the money.
Artificial Analysis places Nemotron 3.5 Super VL first among 67 open-weight models in its 40-billion-to-150-billion-parameter size class. That ranking is specific to the size class: counting each open-weight model once across all sizes, the source material places it around 18th. The model uses a mixture-of-experts design with 120 billion total parameters and 12 billion active, accepts text and images, and has a stated context window of one million tokens.
The listed API rates are $0.20 per million input tokens and $0.575 per million output tokens, with cached input priced at $0.055 per million. Artificial Analysis estimates that its Intelligence Index task costs $0.25. That benchmark cost is not a guaranteed charge for a typical production task: actual bills depend on the amount and type of input, output, and cache use.
Speed figures depend on the measurement. A pre-release DeepInfra endpoint running NVFP4 weights produced about 630 output tokens per second with 100,000-token prompts, according to the source. Across six providers, Artificial Analysis reports a median of 307 tokens per second at that prompt length, with measurements ranging from 106 to 534. Its median output speed on the standard prompt is 196 tokens per second. The figures make provider selection and direct endpoint measurement relevant to deployment decisions.
Infographic 01 · What 25 cents buys
Cheap and fast. Not the most intelligence per dollar.
Nemotron 3.5 Super VL costs $0.25 per Intelligence Index task and scores 26. Several open models with MIT licenses score 13 to 20 points more for the same money or less.
Intelligence Index vs. cost to run it
Up and to the left is better. The dashed line marks Super VL’s $0.25. Hover a dot for license and values.
Not plotted: K2 Horizon 375B A23B (UAE, index 30.5, listed at $0.00) and Nemotron 3.5 Bolt 74B-A9B (index 21.4, no API price).
License matters here. Muse Spark Open (Meta, 38.9) scores higher than Super VL but ships under the Muse Model License, which allows commercial use with restrictions. Super VL’s OpenMDW-1.1 license allows commercial use.
Show the data as a table (20 models)
| Model | Creator | License | Index | Cost / task |
|---|---|---|---|---|
| MiMo-V2.6-Pro | Xiaomi | MIT | 46.3 | $0.13 |
| GLM-5.3 | Z AI | GLM-5.3 License (restricted) | 44.8 | $2.01 |
| Step 5 | StepFun | StepFun Model License (restricted) | 43.7 | $1.03 |
| Kimi K3 | Kimi | Kimi K3 License (restricted) | 43.6 | $2.00 |
| Claude Haiku 5.5 (max) | Anthropic | Proprietary | 43.4 | $0.21 |
| GLM-5.3-Flash | Z AI | MIT | 41.8 | $0.28 |
| Ling 3.1 Flash | InclusionAI | MIT | 41.1 | $1.00 |
| Qwen3.8 Flash Next | Alibaba | Qwen Community License (restricted) | 39.8 | $0.37 |
| DeepSeek V4.1 Flash | DeepSeek | MIT | 39.5 | $0.27 |
| Muse Spark Open | Meta | Muse Model License (restricted) | 38.9 | $0.84 |
| GPT-6 Luna (max) | OpenAI | Proprietary | 38.1 | $0.07 |
| MiMo-V2.6-Flash | Xiaomi | MIT | 37.9 | $0.06 |
| DeepSeek V4 Pro | DeepSeek | MIT | 36.0 | $0.67 |
| Qwen3.8 27B | Alibaba | Apache 2.0 | 33.7 | $1.01 |
| MiniMax-M3 | MiniMax | MiniMax Community License (restricted) | 29.2 | $0.51 |
| Nemotron 3.5 Super VL | NVIDIA | OpenMDW-1.1 | 26.5 | $0.25 |
| Inkling Small | Thinking Machines | Apache 2.0 | 25.7 | $0.09 |
| Nemotron 3 Ultra 550B | NVIDIA | OpenMDW | 22.9 | $0.60 |
| Muse Glimmer | Meta | Apache 2.0 | 17.5 | $0.05 |
| Nemotron 3 Super (previous) | NVIDIA | NVIDIA Open Model License | 12.8 | $1.64 |
Source: Artificial Analysis, Nemotron 3.5 Super VL model page and model data for the comparison models, read 16 Oct 2026; launch analysis (15 Oct 2026). Intelligence Index v4.3.2, ten evaluations. Cost per task = total API cost to run the Intelligence Index divided by its tasks. “Restricted” = Artificial Analysis lists commercial use as allowed with conditions.
Powered by Thorsten Meyer AI
Cost and Speed Depend on Deployment
The model’s reported $0.25 benchmark-task cost combines input, output and cache usage, rather than reflecting the API’s output rate alone. In the supplied breakdown, about 89% of the benchmark cost comes from input, including 69% from cache writes, while output costs less than three cents. Teams processing long documents or repeatedly supplying repository material may therefore see costs driven chiefly by their input pattern.
Speed also has a practical trade-off. The median time to the first token is 0.89 seconds, but the source reports about 11 seconds before the first answer token because the model reasons before responding. A 500-token answer takes about 13.6 seconds end to end in the cited measurements. That profile may suit background processing better than a conversational interface where users expect an immediate answer.
On the same benchmark-cost comparison supplied by the article, several alternatives score higher for a similar or lower estimated task cost. That does not settle which model is best for a particular workload: benchmark scores, license terms, provider performance and the ability to self-host measure different needs. It does mean that speed, license and deployment control are clearer parts of Nemotron’s case than raw score per dollar.
As an affiliate, we earn on qualifying purchases.
A Large Jump From Nemotron 3 Super
The predecessor, Nemotron 3 Super, scored 13 on the Intelligence Index, compared with 26.5 for the new model. Artificial Analysis also estimates its task cost at $1.64, versus $0.25 for Super VL, and reports fewer output tokens per benchmark task for the new model: about 46,000 compared with 114,000. These are benchmark results and estimates, not promises about every user’s workload.
The launch material describes the model weights as released in NVFP4, a four-bit format, and identifies the license as OpenMDW-1.1, with commercial use allowed. The source characterizes Super VL as the highest-scoring permissively licensed U.S. open-weight model. That claim depends on how permissive licensing and U.S. origin are defined. It also distinguishes the model from Meta’s Muse Spark Open, which the source says scored 39 but has a license permitting commercial use with restrictions.
The supplied task-level results point to uneven capabilities. Artificial Analysis reports a GDPval-AA score of 1,162 Elo, up from 490 for Nemotron 3 Super, and an AA-Briefcase score of 999 Elo. AutomationBench-AA is reported at 14.8%, compared with 3.0% for Nemotron 3 Ultra. On AA-Omniscience, the reported hallucination rate fell from 87% to 36% versus the predecessor. The source also cautions that the automation result remains low in absolute terms.
Endpoint Results and Real-World Costs
The source figures do not establish how Nemotron 3.5 Super VL will perform on a reader’s own workload. Provider speeds span 106 to 534 tokens per second in the reported cross-provider distribution, and the 630-token figure came from a single pre-release endpoint. The available material does not identify whether those exact conditions will persist in general availability or across different hardware and prompt mixes.
The $0.25 estimate is tied to Artificial Analysis’s benchmark task. The source gives a breakdown in which input and cache activity account for most of that cost, but actual production usage can differ. It also does not provide enough detail here to establish how the reported hallucination rate translates to every application, or to compare reliability across a reader’s specific tasks.
License suitability also requires attention to the terms themselves. The source says commercial use is allowed under OpenMDW-1.1, but does not reproduce the complete license or explain every condition. Teams considering deployment should review the license and confirm that the model’s provenance, hosting arrangement and operational requirements meet their needs.
Test the Intended Provider and Workload
The immediate next step for teams evaluating the model is to measure it on the provider and hardware they plan to use, with representative prompt lengths, cache behavior and output sizes. That can show whether published latency and speed figures match their application and whether input-heavy tasks fit the estimated budget.
Evaluation should also cover the relevant task-level performance, including document work, automation and error rates, rather than relying only on the Intelligence Index average. Teams considering self-hosting can assess the released NVFP4 weights against their hardware and operating needs, then review the complete OpenMDW-1.1 terms before commercial deployment.
The supplied material identifies no upcoming independent milestone or scheduled update. Real-world costs, provider consistency and performance on specific workloads remain to be established through deployment evidence.
Key Questions
What does the $0.25 task figure measure?
It is Artificial Analysis’s estimated cost for one Intelligence Index task. It is not a fixed price for every production request; token volume and cache use affect the bill.
Is 630 tokens per second the expected speed?
No single speed applies to every provider. The source reports about 630 tokens per second on one pre-release DeepInfra endpoint with 100,000-token prompts, compared with a 307-token-per-second median across six providers at that prompt length.
How does it compare with Nemotron 3 Super?
Artificial Analysis reports a score increase from 13 to 26.5 on the Intelligence Index and a benchmark-task cost decrease from $1.64 to $0.25. Those are benchmark results and cost estimates, not guaranteed outcomes for other workloads.
Does the license permit commercial use?
The source identifies the license as OpenMDW-1.1 and says commercial use is allowed. The full license terms and any conditions should be reviewed before deployment.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
