🔍 Read the full analysis: Nemotron 3.5 Super VL’s Cost, Speed, And Score In Focus on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
NVIDIA released Nemotron 3.5 Super VL on October 15, 2026. Artificial Analysis gave it an Intelligence Index score of 26. The model is priced below the median for its size class and can be very fast on some endpoints, but its speed varies by provider and several alternatives score higher for similar or lower task costs.
NVIDIA released Nemotron 3.5 Super VL on October 15, with Artificial Analysis reporting an Intelligence Index score of 26 and a cost of about $0.25 per index task. The 120-billion-parameter open-weight model is built for text and image input, and its reported speed, commercial-use license and option for on-premises deployment may interest teams weighing operational control against benchmark performance.
Artificial Analysis puts the model at 26.5 on its Intelligence Index, rounded to 26 in the headline figures, twice the score of its predecessor, Nemotron 3 Super, which scored 13. It ranks first among 67 open-weight models in the 40-billion-to-150-billion-parameter size class. That is a class-specific result, not a top ranking across all open-weight models; the source estimates its overall position at about 18th.
The model has 120 billion total parameters and 12 billion active parameters in a mixture-of-experts design, and a stated context window of one million tokens. NVIDIA released weights in NVFP4, a 4-bit format. The listed API prices are $0.20 per million input tokens and $0.575 per million output tokens; cached input is listed at $0.055 per million tokens. Artificial Analysis estimates a full index task costs $0.25, compared with $1.64 for Nemotron 3 Super.
Speed figures depend on the endpoint. A pre-release DeepInfra endpoint serving NVFP4 weights produced about 630 output tokens per second with prompts of 100,000 tokens, according to the source’s account of Artificial Analysis measurements. Across six providers, the model page reports median output speeds of 196 tokens per second on standard prompts and 307 with 100,000-token prompts. At that long-prompt setting, the reported range was 106 to 534 tokens per second. Median time to first token was 0.89 seconds, but the model’s reasoning phase can mean about 11 seconds before the first answer token.
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
Speed and Control Versus Score
For teams building high-volume pipelines or background agents, lower task cost, short time to first token and the option to run model weights on their own NVIDIA hardware could matter as much as an aggregate benchmark. The listed license, OpenMDW-1.1, allows commercial use, according to the source. That combination may help organizations that need deployment control or have specific licensing requirements.
But the figures do not establish Nemotron 3.5 Super VL as the best-value model for general capability. The source compares it with models that score higher at similar or lower estimated task costs: Xiaomi’s MiMo-V2.6-Pro scores 46 at $0.13 per task, while MiMo-V2.6-Flash scores 38 at $0.06. DeepSeek V4.1 Flash scores 39 at $0.27, and GLM-5.3-Flash scores 42 at $0.28. These are benchmark and cost comparisons, not guarantees of performance on a particular workload.
The result is a trade-off rather than a single winner. Speed, deployment options and license terms are distinct from raw benchmark score. Teams should also account for the roughly 11-second reasoning interval reported before an answer begins, as well as provider-specific speed and their own input-token use.
Products Worth Considering
Trezor Safe 5 Crypto Hardware Wallet with Color Touchscreen, Black Graphite
Ledger Flex + Folio Crypto Hardware Wallet
Ledger Nano Gen5 Crypto Wallet, Black
As an affiliate, we earn on qualifying purchases.
How the Benchmark Compares
Artificial Analysis measures models across an Intelligence Index and task-specific evaluations. The headline score summarizes performance across that framework; it does not mean the model will perform equally well on every task. The reported first-place ranking is limited to the stated 40B-to-150B open-weight size class. The source says the model ranks about 18th when comparing open-weight models overall.
Cost per index task depends on the evaluation’s token use, not only the published per-token rates. Artificial Analysis’ launch figures put Nemotron 3.5 Super VL’s task output at about 46,000 tokens, down from roughly 114,000 for Nemotron 3 Super. The source estimates that about 89% of the new model’s $0.25 task cost comes from input, with cache writes accounting for 69%; output costs less than three cents. For agents that repeatedly process long documents or code repositories, input and caching patterns can shape the bill.
The model also showed gains in specific reported tests. Its GDPval-AA score, covering economically useful tasks, was 1,162 Elo, compared with 490 for its predecessor. AutomationBench-AA was 14.8%, versus 3.0% for Nemotron 3 Ultra. On AA-Omniscience, the reported hallucination rate fell from 87% for Nemotron 3 Super to 36%. Those results describe named evaluations; they do not prove a particular error rate or success level in production.
“Your bill will follow your input profile, not the list price for output.”
— Thorsten Meyer, describing the cost breakdown
Provider Results and Real-World Fit
The cited speed results are not a single guaranteed operating rate. The highest figure came from one pre-release endpoint, while the six-provider median and the wide reported range show that hosting choice can change results. The source does not provide enough detail here to establish how those measurements compare across hardware configurations, service conditions or later production deployments.
Benchmark scores and estimated evaluation costs also do not settle how the model will perform on an organization’s own tasks. The supplied material does not specify a full production test, independent replication of every figure, or the cost of operating self-hosted weights, including hardware and maintenance. The reported license permits commercial use, but teams still need to review its terms for their intended deployment.
One comparison also depends on what a buyer means by “permissive.” The source says Nemotron 3.5 Super VL leads among permissively licensed US open-weight models, while Meta’s Muse Spark Open scored 39 under a license described as permitting commercial use with restrictions. The practical distinction depends on a team’s legal requirements; the source does not provide a legal analysis of either license.
Test the Intended Deployment
The immediate next step for prospective users is to test the model on the provider and workload they expect to use. That means measuring response latency, throughput, input and cache-token use, accuracy on representative tasks, and total cost—not assuming the fastest endpoint result will carry over to another service or self-hosted setup.
Further provider measurements and independent production reports could clarify how consistent the speed and cost figures are beyond the launch-day evaluations. Teams considering self-hosting will also need to check whether their available NVIDIA hardware can serve the released weights at an acceptable rate and cost. No future benchmark update or deployment milestone is specified in the source material.
Key Questions
What score did Nemotron 3.5 Super VL receive?
Artificial Analysis reported an Intelligence Index score of 26.5, commonly rounded to 26. It ranked first among 67 open-weight models in the 40B-to-150B size class, but the source estimates its position across all open-weight models at about 18th.
How much does a benchmark task cost?
Artificial Analysis estimates $0.25 per Intelligence Index task. That figure reflects the evaluation’s token use and is not a universal price for every customer workload. Listed API rates are $0.20 per million input tokens and $0.575 per million output tokens, with cached input at $0.055 per million tokens.
How fast is the model?
A pre-release DeepInfra endpoint reached about 630 output tokens per second on 100,000-token prompts. Across six providers, the reported median was 307 tokens per second at that prompt length, with a range of 106 to 534. Results depend on the provider and setup.
Can companies use the model commercially?
The source identifies its license as OpenMDW-1.1 and says commercial use is allowed. Organizations should review the license terms for their particular product and deployment before relying on that description.
Does the benchmark make it the best-value model?
Not by the comparisons in the source. Several alternatives have higher reported scores at similar or lower estimated task costs. Nemotron 3.5 Super VL’s case instead includes speed on some endpoints, commercial-use licensing and the option to run released weights on NVIDIA hardware.
Source: ThorstenMeyerAI.com
