The chip maker just decided it wants to be a model maker too, and it's giving the weights away for free.
The Summary
- Nvidia released Nemotron 3.5 Super VL, a new vision-language model with open weights that directly challenges Claude and GPT
- Adobe is integrating Nemotron into its Brand Concierge for enterprise customer engagement with built-in security and compliance
- The strategy emphasizes efficiency and transparency over proprietary lock-in, but performance claims still need independent validation
The Signal
Nvidia isn't content selling the shovels anymore. The company's new Nemotron 3.5 Super VL model positions the hardware giant as a direct competitor to OpenAI and Anthropic in the foundation model space. The twist: they're releasing it with open weights, meaning anyone can download, modify, and deploy it without licensing restrictions.
This is the same playbook Meta used with Llama, but with Nvidia's GPU infrastructure advantage baked in. The model handles both vision and language tasks, the table stakes for any serious AI product in 2026. If you're building customer service agents, content moderation systems, or visual search tools, you now have another option that doesn't require sending your data to San Francisco.
"Open weights and efficiency could reshape AI competition away from proprietary lock-in."
Adobe's integration into Brand Concierge shows what enterprise adoption looks like when a chip maker releases models. Brand Concierge is Adobe's AI-powered customer engagement platform, where companies manage interactions across channels while maintaining brand consistency. Adding Nemotron means enterprises get:
- GPU-optimized inference that runs faster on Nvidia hardware they already own
- Security and compliance features designed for regulated industries
- Transparency into model behavior, critical for brand safety and legal review
The Adobe deal matters because it's not a demo or a partnership announcement with vague timelines. It's shipping product integration from a company that serves Fortune 500 marketing departments. That's real distribution, not a blog post.
But there's a credibility gap here. Nvidia's performance claims need independent validation. Every model release comes with cherry-picked benchmarks that make it look like the second coming. Until researchers outside Nvidia run their own tests, treat the "takes aim at Claude and GPT" framing as aspirational. The open weights strategy helps here—third parties can actually evaluate it instead of taking marketing slides at face value.
The Implication
Watch for two things. First, whether other enterprise software vendors follow Adobe's lead and integrate Nemotron as an alternative to OpenAI or Anthropic APIs. If you're building agents that need vision capabilities, open weights mean you can fine-tune for your specific use case without waiting for OpenAI to add features or Anthropic to lower pricing.
Second, this is Nvidia hedging against a world where foundation models become commoditized. If every model performs roughly the same, the value shifts back to infrastructure, fine-tuning tools, and deployment platforms. Nvidia wins that game. Releasing competitive models with open weights accelerates commoditization while keeping everyone dependent on H100s and whatever comes next.