MiMo-V2.6-Pro

MiMo-V2.6-Pro is Xiaomi's flagship in the MiMo-V2.6 series, positioned as the company's most capable model to date and built as a natively omnimodal foundation that handles frontier reasoning across multiple input types. The series was released and open-sourced on September 22, 2026, accompanied by a lighter MiMo-V2.6-Flash variant tuned for efficiency and a MiMo-V2.6-Pro-UltraSpeed variant that Xiaomi states delivers up to 20× faster output at equivalent quality for latency-sensitive deployments. Development centered on scaling RL compute on verifiable, complex tasks—what Xiaomi describes as their RSI (recursive self-improvement) path—with six days of public "Live RL" training that built on roughly half a year of prior foundational research and engineering iteration.

In Xiaomi's own benchmark reporting, MiMo-V2.6-Pro posts strong results across agentic, coding, and general intelligence evaluations, reaching 46 on the Artificial Analysis Intelligence Index—reported by Xiaomi as the highest among open-source models, surpassing Kimi K3 and Qwen3.8 Max, though still trailing closed-source leaders such as Claude Fable 5.1 and GPT-6 Astra. On coding-focused suites the model scores 63.2 on MiMo Code Bench and 71.9 on DeepSWE v1.1, while agentic benchmarks include 53.1 on Automation Bench v1.0.6, 31.6 on Agents' Last Exam, 76.9 on Toolathlon-verified, and 1673 on GDPVal 2.1. The combination of open weights, omnimodal inputs, and tool/attachment-friendly capabilities makes the model well suited to professional workflows that need strong reasoning and agentic execution without closed-source dependencies.

Tempr Gatewayxiaomi-mimo/mimo-v2.6-promimo

Open provider page Provider docs

Quick Info

Provider: Tempr Gateway
Model key: xiaomi-mimo/mimo-v2.6-pro
Release date: Sep 22, 2026
Last updated: Sep 22, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.435
Output token cost: $0.87

Limits

Output tokens: 131,072 tokens
Context window: 1,048,576 tokens

Transparent token rates

Compare MiMo-V2.6-Pro pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about MiMo-V2.6-Pro

Tempr Gateway

Coverage

Xiaomi released MiMo-V2.6-Pro on September 21, 2026, debuting as the top open-weight model on Artificial Analysis' Intelligence Index with a score of 46. That ties Grok 4.7 and surpasses Grok 4.6 (44), Gemini 3.8 Flash (41), and DeepSeek V4.1 Flash (39) and Pro (36), according to VentureBeat's reporting. The model is MIT-licensed and openly available on Hugging Face for download and fine-tuning. MiMo-V2.6-Pro runs a 1-million-token multimodal context window handling text, images, audio, and video, with output speed around 134 tokens per second. Xiaomi's own API prices it at $0.435 per million input tokens and $0.87 per million output tokens, roughly $0.13 per Artificial Analysis Intelligence Index task. A companion MiMo-V2.6-Flash variant shares the same context window at $0.14/$0.28 per million tokens.

Tempr Gateway

Coverage

MiMo-V2.6-Pro carries a 1-million-token context window and approximately 1.02T parameters, with full Pro weights released under the MIT license on Hugging Face for free download, according to a Kingy.ai integration guide. A distilled 9B variant, MiMo-V2.6-Distill-Qwen-9B, is also available as a much weaker open derivative. Xiaomi's MiMo Studio provides a browser-based chat demo requiring account sign-in. Xiaomi's API lists MiMo-V2.6-Pro at $0.435 per million input tokens and $0.87 per million output tokens, positioning it well below Claude and GPT rates. Developers can also reach the model through Xiaomi's OpenAI- and Anthropic-compatible API, OpenRouter listings, and Xiaomi's Token Plan starting at $6 per month. The guide is based on Xiaomi's launch post and API documentation, plus OpenCode and OpenRouter references.