VercelZachary Chen1 min readrelease notesintermediate
MiMo V2.6 models now available on AI Gateway
Summary
Vercel AI Gateway now offers three Xiaomi MiMo V2.6 models (Pro, Flash, Pro UltraSpeed) with up to 1 M token context, 1 T total parameters (42 B active per token) for Pro, 309 B total (15 B active) for Flash, and a 20× speed variant. The models support multimodal inputs and can be used via Vercel’s unified API and CLI for coding agents.
- MiMo V2.6 Pro: 1.02 T total params, 42 B active per token – targets complex software‑engineering and long‑running agents.
- MiMo V2.6 Flash: 309 B total params, 15 B active per token – more compute‑efficient for everyday multimodal automation.
- MiMo V2.6 Pro UltraSpeed: same capabilities as Pro but up to 20× faster output for latency‑sensitive workloads.
- All models expose a 1 M token context window and can emit up to 128 K output tokens, enabling long‑repo analysis and multi‑session agents.
The addition of high‑parameter, multimodal MiMo models to a managed gateway lowers the operational friction for teams building AI‑powered coding assistants, especially when long context windows and low latency are required. However, the post is largely a product announcement with minimal engineerin…
4/10
