New AI Models Reach Momen and Two Developer Gateways.

Three platforms added AI models built for more than text generation this week. Momen added two models for long-running agent work, while Netlify and Vercel expanded their AI gateways with models for routing, coding, reasoning and tool use.

The releases cover three different patterns. Momen’s models can run a sequence of actions and return data an app can use. Netlify’s addition attaches confidence scores to several answers from one call. Vercel’s new model family handles long contexts and multiple media types.

TypeSafe AI’s Jev is now available through Netlify AI Gateway, according to Netlify’s LinkedIn post. In the walkthrough, Jev processes a support ticket about Stripe Connect and returns the responsible team, a customer frustration score and the probability of genuine urgency from one call, with the questions evaluated in parallel.

Results below 0.5 escalate to a heavier reasoning model, leaving Jev to process the easier cases before a more expensive model is called.

GPT 6 Astra and Gemini 3.8 Flash have joined Momen’s model list, according to its LinkedIn post. Both models support long-horizon work in which an agent plans, calls a tool, reads the result and continues through the task.

Tool calling and structured output let either model call a Momen Actionflow, read a table and return a typed object that the app can act on, rather than stopping at a text response. The addition follows Momen’s earlier expansion of AI Agent model selection across nine models.

MiMo V2.6 Pro, MiMo V2.6 Flash and MiMo V2.6 Pro UltraSpeed from Xiaomi are now available through AI Gateway, according to the Vercel changelog. The family combines coding, reasoning and tool use with native understanding of text, images, audio and video, plus a 1M-token context and outputs of up to 128K tokens.

MiMo V2.6 Pro has 1.02T total parameters with 42B activated per token, while Flash has 309B total parameters with 15B activated per token. UltraSpeed provides output up to 20 times faster, and builders can integrate the models through the AI SDK or configure a coding agent with the Vercel CLI.

What are GPT 6 Astra and Gemini 3.8 Flash in Momen?

They are models for long-horizon agent work in Momen’s model list. They can plan, call tools such as an Actionflow, read results and return structured objects that an app can use.

Which plans include these new AI models?

All three announcements describe the models as available now through their respective model lists or gateways. The sources do not name required paid plans or prices.

How do these new AI models compare?

Momen’s additions focus on long-running tool sequences and structured output. Jev handles several evaluations in parallel and adds confidence values for routing, while MiMo V2.6 combines coding, reasoning, tool use and multimodal understanding with a 1M-token context.