🔍 Read the full analysis: Does Nemotron 3.5 Super VL’s Score Match Its Speed? on ThorstenMeyerAI.com
Get the latest gadgets 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 Oct. 15, 2026. Artificial Analysis scored it 26, while a pre-release endpoint reached about 630 output tokens per second on long prompts; median speeds across six providers were lower. Its $0.25 cost per index task and permissive commercial-use license may appeal to teams prioritizing deployment control, but several alternatives score higher or cost less.
The model has 120 billion total parameters, with 12 billion active in its mixture-of-experts design, and a stated one-million-token context window. It accepts text and images and produces text. NVIDIA released NVFP4 weights, and the model is offered under the OpenMDW-1.1 license, which the source says permits commercial use. Artificial Analysis lists API prices of $0.20 per million input tokens and $0.575 per million output tokens; cached input is $0.055 per million.
Artificial Analysis assigns Super VL a $0.25 cost per Intelligence Index task, compared with $1.64 for Nemotron 3 Super. The source attributes the lower task cost partly to fewer generated tokens: about 46,000 per index task, versus 114,000 for the predecessor. In the reported cost breakdown, input accounts for about 89% of the task bill, with cache writes alone representing 69%; output costs less than three cents. These figures describe the benchmark task, and actual bills will depend on usage and caching.
Speed results depend on where the model runs. A pre-release DeepInfra endpoint serving NVFP4 weights produced about 630 output tokens per second with 100,000-token prompts. Across six providers, Artificial Analysis lists a median of 307 tokens per second for that prompt length, with a range from 106 at the fifth percentile to 534 at the 95th. Median first-token latency is 0.89 seconds, but the source estimates about 11 seconds before the first answer token when reasoning is included, and about 13.6 seconds for a 500-token answer.
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 Comes With a Trade-Off
For teams running large document or codebase workflows, Super VL combines a long context window, low benchmark-task cost and the option to use released weights on NVIDIA hardware. Those features can matter when a team needs control over deployment or has workloads that generate many cached-input tokens. The license may also fit teams whose legal requirements favor permissively licensed US open-weight models.
Its benchmark position is less compelling if the goal is maximum measured intelligence per dollar. The source lists MiMo-V2.6-Pro at 46 points for $0.13 per task, MiMo-V2.6-Flash at 38 for $0.06, and Inkling Small at the same 26 points for $0.09. DeepSeek V4.1 Flash scores 39 for $0.27, while GLM-5.3-Flash scores 42 for $0.28. These comparisons come from the supplied source’s Artificial Analysis figures; they do not establish equal performance on every workload.
Operationally, an 11-second reasoning delay can be acceptable in a background pipeline yet conspicuous in a live chat. The reported throughput also comes from different provider measurements, so it is not a guarantee for a particular deployment. Teams evaluating the model should consider their own endpoint, prompt lengths, latency needs, licensing requirements and input-to-output mix.
As an affiliate, we earn on qualifying purchases.
How the Benchmark Ranks It
Artificial Analysis compares open-weight models in size bands. Super VL is ranked No. 1 among 67 models in its 40-billion-to-150-billion-parameter class, according to the source. Across all open-weight models, it ranks about 18th when each model is counted once. The class ranking and overall ranking describe different comparison groups.
The Intelligence Index score is reported as 26, with a more precise figure of 26.5 in the supplied spec sheet. Its predecessor, Nemotron 3 Super, scored 13. The source also reports gains on selected evaluations: GDPval-AA, a measure of economically useful work, rose from 490 to 1,162 Elo; AA-Briefcase reached 999, compared with 876 for Nemotron 3 Ultra. AutomationBench-AA measured 14.8%, versus 3.0% for Ultra.
On AA-Omniscience, the reported hallucination rate fell from 87% for Nemotron 3 Super to 36% for Super VL. The source presents this as a benchmark result, not as a guarantee of factual accuracy in all uses. It also says Super VL leads among permissively licensed US open-weight models, while Meta’s Muse Spark Open scored 39 under a license described as commercially usable with restrictions.
“Super VL is a very good engine and an average brain.”
— Thorsten Meyer, ThorstenMeyerAI.com
As an affiliate, we earn on qualifying purchases.
Results Depend on Deployment
The supplied figures do not establish what throughput a buyer will receive from a specific provider or self-hosted setup. The 630-token-per-second result came from one pre-release endpoint; provider measurements ranged from 106 to 534 tokens per second at the 5th and 95th percentiles. Hardware, serving configuration and workload may affect results, but their effects are not detailed here.
It is also unclear how closely the index and task-cost measurements predict performance on a team’s own tasks. The source provides selected benchmark results, not a broad account of real-world production outcomes. The 36% hallucination rate is specific to AA-Omniscience, and should not be read as an overall error rate.
The license comparison is summarized rather than accompanied by the full license text or legal analysis. Teams will need to review the actual terms for their use case. Pricing and provider availability may also change after launch; the source gives launch-time figures and does not specify how long they will remain in effect.
As an affiliate, we earn on qualifying purchases.
Test the Intended Endpoint
The immediate next step for prospective users is to measure the provider or hardware setup they plan to use. A useful evaluation should match their expected prompt lengths and output sizes, and track latency, throughput and input costs alongside answer quality. This is particularly relevant for document-heavy agents, where cached input made up much of the reported benchmark bill.
Further provider measurements, post-release serving results and evaluations on task-specific workloads could clarify whether Super VL’s speed and deployment options offset its lower general benchmark score for particular teams. The supplied material does not give a schedule for additional benchmarks or model updates, so those developments remain unannounced.
As an affiliate, we earn on qualifying purchases.
Key Questions
What score did Nemotron 3.5 Super VL receive?
Artificial Analysis reported an Intelligence Index score of 26; the supplied spec sheet gives a more precise value of 26.5. Its predecessor scored 13.
Is 630 tokens per second the typical speed?
No. That figure came from one pre-release DeepInfra endpoint using NVFP4 weights and 100,000-token prompts. Across six providers, the reported median for that prompt length was 307 tokens per second.
How much does it cost per benchmark task?
Artificial Analysis lists a cost of $0.25 per Intelligence Index task. Actual costs depend on the workload, provider and token usage; the reported benchmark breakdown was dominated by input and cache-write costs.
Does its license allow commercial use?
The supplied source says the OpenMDW-1.1 license allows commercial use. Users should review the full license terms for their intended deployment.
Does Super VL beat other open models on score per dollar?
Not in the comparisons provided. Several listed open models scored higher for similar or lower task costs, and Inkling Small matched its score of 26 at a lower reported cost. Model choice may also depend on speed, licensing and deployment needs.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
