Vercel AI Gateway adds Nemotron 3 Ultra

// 3h agoPRODUCT UPDATE

Vercel AI Gateway adds Nemotron 3 Ultra

Vercel has integrated NVIDIA's Nemotron 3 Ultra model (nvidia/nemotron-3-ultra-550b-a55b) into its AI Gateway, offering developers access to the 550B parameter open-weight model. The hybrid Transformer-Mamba model is optimized for agentic coding and reasoning, and is available at a 30% lower cost than comparable frontier models.

// ANALYSIS

Offering high-capacity open-weight models at reduced prices within unified APIs is key to accelerating the adoption of complex agentic workflows.

* Cost Reduction for Agents: Agentic loops consume large amounts of tokens, making a 30% cost savings crucial for production viability.

* Hybrid Architecture: The Transformer-Mamba hybrid architecture enables efficient scaling and fast inference for longer-context developer tasks.

* Developer Abstraction: By hosting Nemotron 3 Ultra, Vercel reinforces its AI Gateway as a versatile, provider-agnostic layer for modern web applications.

// TAGS

vercelnvidianemotronai-gatewayopen-weightllmagent

DISCOVERED

3h ago

2026-06-04

PUBLISHED

3h ago

2026-06-04

RELEVANCE

8/ 10

AUTHOR

vercel_dev

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

NEWS13m ago

Anthropic seeks Claude Code performance PM

Anthropic is hiring a Product Manager for Claude Code to focus on model performance and agentic evaluations. The position, shared in an announcement retweeted by Lydia Hallie, calls for candidates with experience designing and implementing evaluations for agentic systems.

UPDATE48m ago

OpenAI adds web app builder to ChatGPT

Sam Altman announced a new ChatGPT feature enabling users to build and publish web applications instantly, comparing the experience to Apple's classic HyperCard platform. By combining AI-driven code generation with direct web hosting and publishing capabilities, this update allows non-programmers and creators to bring interactive apps to life and share them with a single click.

LAUNCH1h ago

DigitalOcean launches Model Evaluations preview

DigitalOcean Model Evaluations has launched in public preview within the DigitalOcean Inference Engine. The tool lets developers test and compare LLMs, Hugging Face models, and routing configurations on custom datasets to optimize cost, latency, and performance.