Deepinfra
Day 45 after signup
November at DeepInfra: New Nemotron models, API scopes and more
Four NVIDIA Nemotron 3.5 models, three other new models, API key scopes, and GTC Berlin.
Why it works
- Organize a broad product roundup into clearly labeled sections so readers can scan for the updates most relevant to them.
- Give each technical update a brief explanation before linking to deeper details, keeping the overview useful without making it overly long.
- Close with a short human note from the team after the product news to add a personal touch without changing the email's informative focus.
How it is built
- Monthly introduction
- Featured model release
- Additional model updates
Email copy
November 2026 Here’s what's new on DeepInfra: four NVIDIA Nemotron 3.5 models, other trendy models, scopes for API keys, and a note from GTC Berlin. NVIDIA Nemotron 3.5 Four Nemotron 3.5 models added. NVIDIA released several Nemotron 3.5 models in October, and each was available on DeepInfra on its release day. Together they cover image and document understanding, retrieval, and speech. Super VL is a multimodal model that takes text, images, documents, and video as input and supports reasoning and tool calling. Embed and Rerank are retrieval models: Embed indexes text and document images in the same search space, and Rerank scores retrieved results against a query. Transcribe converts up to an hour of audio into timestamped text in 40 languages. Super VL release post → Embed and Rerank release post → Transcribe release post → Other new models Three more models in the catalog. DeepSeek-V4.1-Flash-1017 is an update to DeepSeek-V4.1-Flash. DeepSeek reports higher scores on reasoning, coding, and agent benchmarks, with shorter outputs per task. GLM-5.3-Flash from Z.ai is the first model in the GLM-5 series to accept images as well as text. It has a context window of about 1M tokens and is aimed at coding and agent tasks. Kimi-K3.1-Pro from Moonshot AI accepts text and image input, has a context window of about 1M tokens, and is aimed at coding, research, and agent tasks. Browse the model catalog → API key scopes API keys can now be scoped. You can now limit what an API key is allowed to do. Instead of full access to the account, a key can be restricted to specific areas, such as running inference, reading billing information, or managing sandboxes and agents. View the list of scopes → GTC Berlin Thank you for stopping by. Thank you to everyone who visited us at our booth at NVIDIA GTC Berlin. It was great to meet you in person and hear how you’re putting open models to work. That’s all for now. — The DeepInfra team Models · Docs · Pricing · · You are receiving this because you signed up at deepinfra.com. Deep Infra Inc · 2625 Middlefield Rd PMB 460 · Palo Alto, CA 94306-2516 · USA Unsubscribe · Update preferences