Vercel AI Updates: October 14, 2026

1. Google’s Gemini 4 Argon Available on AI Gateway

Vercel. Vercel added Google’s Gemini 4 Argon to AI Gateway under the model ID google/gemini-4-argon, with a 1M-token context window, text, image, PDF, and video input, and tool calling. Vercel positions the model for coding, agentic work, and complex reasoning, and notes that it uses low, medium, and high thinking levels rather than thinking budgets. Developers can call it through the AI SDK’s streamText, the OpenAI-compatible Chat Completions endpoint at https://ai-gateway.vercel.sh/v1, the Responses and Anthropic Messages APIs, or Claude Code after running npx vercel ai-gateway setup. AI Gateway passes through provider pricing with no markup and no platform fee on inference, including bring-your-own-key requests. Source

2. Microsoft MAI-Image-2.6 and MAI-Image-2.6-Flash Come to AI Gateway

Vercel. Vercel added two Microsoft image models to AI Gateway: microsoft/mai-image-2.6, aimed at final assets where quality matters more than speed, and microsoft/mai-image-2.6-flash, a faster and cheaper option for iterating on prompts and variations. Both support photorealistic generation, text rendering in images, targeted edits to products, portraits, and layouts, up to five reference images, and output sizes up to 1536 x 1536. The models can be called through the AI SDK’s generateImage, the OpenAI-compatible images.generate and images.edit endpoints, or the AI CLI with npx ai-cli image, using a single gateway key with no separate Microsoft account. Source

3. Fish Audio’s Drama 3 Text-to-Speech Model in Preview on AI Gateway

Vercel. Vercel added Fish Audio’s Drama 3 text-to-speech model to AI Gateway in preview under the model ID fish-audio/drama-3. The model takes natural-language directions in square brackets for emotion and pacing, inline sounds and pauses, and word-level tags such as <whisper> and <emphasis>, and it supports multi-speaker dialogue with a separate voice per character and directions written in a different language from the spoken text. Developers call it through the AI SDK’s generateSpeech function with text, a voice ID, and an output format, and Vercel is offering 80% off for 30 days through November 12. Source