Meta and Nvidia both released large open-weight AI models on October 15, 2026. Meta shipped Muse Spark Open, and Nvidia shipped Nemotron 3.5 Super VL. Both use mixture-of-experts designs and a 1M-token context window. Here is what each model offers and what it means for teams that build with AI.
What Did Meta Release With Muse Spark Open?
Muse Spark Open is an 816B-parameter mixture-of-experts model with 42B active parameters and a 1M-token context window. Meta’s Superintelligence Lab released it under the Muse Model License, which allows free use for many developers. Details come from Meta’s research blog.
Meta says the model runs on two to four paired M5 Ultra Mac Studios through MLX. It also calls Muse Spark Open “the most capable US open-weight model to date.” That claim is Meta’s own and has no independent benchmark yet.
What Is Nvidia Nemotron 3.5 Super VL?
Nemotron 3.5 Super VL is a 120B-parameter mixture-of-experts model with 12B active parameters and a 1M-token context window. Nvidia released the weights on October 15 under the OpenMDW license. It also shipped two smaller variants: Bolt 74B-A9B and E4B for edge devices.
Nvidia claims up to 5x higher throughput than comparable open models on a single GPU. It also says the model beats Claude Opus 5 and GPT-6.1 Sol on five specialized tasks. VentureBeat covered the benchmark claims. Treat them as vendor results until third parties repeat them.
How Do the Two Open-Weight Models Compare?
Muse Spark Open is larger, while Nemotron 3.5 Super VL is lighter and built for throughput. Both give developers a 1M-token window and open weights.
| Model | Total / Active Parameters | Context Window | License |
|---|---|---|---|
| Meta Muse Spark Open | 816B / 42B | 1M tokens | Muse Model License |
| Nvidia Nemotron 3.5 Super VL | 120B / 12B | 1M tokens | OpenMDW |
| Nvidia Nemotron Bolt | 74B / 9B | Not stated | OpenMDW |
Why Do Open-Weight Releases Matter for Business Teams?
Open weights let a company run a model on its own hardware, which keeps data in-house and cuts per-token fees. Teams that handle private code or customer records gain the most from this setup.
Open models also pressure closed vendors on price. Developers choosing a coding tool can review our guide to the best AI coding assistants. Teams building autonomous workflows can compare options in our list of the best AI agents. Leaders weighing rollout risk can read our AI governance guide.
What Else Happened in AI on October 15?
The FTC sent separate 50-page civil investigative demands to OpenAI and Anthropic, and DeepSeek neared a roughly $14B funding round.
Axios first reported the FTC demands. They ask for records on security incidents, safeguards, marketing materials, and privacy policies. The commission plans to subpoena more AI firms. Seoul Economic Daily reported that DeepSeek is closing a 100-billion-yuan round at a 500-billion-yuan valuation to fund a data center with 160,000+ Huawei Ascend chips.
The Bottom Line
Open-weight models now match closed ones on context size, and two big vendors shipped them on the same day. Test both on your own tasks before you commit, since the benchmark claims come from the vendors. We will update this page as independent results arrive. See more in our AI for business guide.
Frequently Asked Questions
What is Meta Muse Spark Open?
Muse Spark Open is Meta’s open-weight mixture-of-experts model with 816B total parameters, 42B active parameters, and a 1M-token context window, released October 15, 2026.
Is Muse Spark Open free to use?
The Muse Model License permits free use for many developers. Read the license terms before you use it in a commercial product.
What is Nvidia Nemotron 3.5 Super VL?
It is a 120B-parameter mixture-of-experts model with 12B active parameters, a 1M-token context window, and open weights under the OpenMDW license.
Can Muse Spark Open run on a desktop?
Meta says it runs on two to four paired M5 Ultra Mac Studios through MLX.
Are the benchmark claims verified?
No. Meta and Nvidia published the claims themselves. Independent tests have not yet confirmed them.
Related Articles
David Austin is a technology writer and software analyst at DeployHyre, where he covers AI tools, SaaS platforms, cloud hosting, and business automation. He focuses on hands-on comparisons of pricing, features, and real-world performance so teams can pick the right software with confidence.