MiMo-V2.6-Flash

MiMo-V2.6-Flash is one of two natively omnimodal models Xiaomi released and open-sourced as part of the MiMo-V2.6 series on September 22, 2026. Xiaomi positions the Flash variant as the option that strikes the best balance between intelligence, efficiency, and cost, sitting beneath the flagship MiMo-V2.6-Pro while offering broader accessibility for production use. The series itself is framed around scaling reinforcement learning compute on verifiable, complex tasks, with the explicit goal of expanding the capability frontier through exploration and feedback rather than relying solely on pretraining scale.

In Xiaomi's own benchmark charts, MiMo-V2.6-Flash shows particular strength on coding and software engineering tasks, scoring 67.9 on DeepSWE v1.1, 26.0 on ProgramBench, and 61.2 on the in-house MiMo Code Bench, placing it ahead of the prior MiMo-V2.5-Pro across these evaluations and competitive with larger peer models on agentic workloads. Third-party aggregator rankings place the model in the top tier for coding (22 of 273) and tool calling, with average positioning on reasoning and vision, reflecting a profile well suited for multi-agent collaboration, tool-driven workflows, and large-scale coding projects rather than open-ended chat.

Xiaomi Token Plan (Europe)mimo-v2.6-flashmimo

Open provider page Provider docs

Quick Info

Provider: Xiaomi Token Plan (Europe)
Model key: mimo-v2.6-flash
Release date: Sep 22, 2026
Last updated: Sep 22, 2026
Input modalities
Output modalities
Capabilities

Cost

A provider subscription or plan supersedes token-based pricing for this model.

Limits

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

Latest news about MiMo-V2.6-Flash

Xiaomi Token Plan (Europe)

Official sourceDocumentation

Xiaomi's official MiMo documentation page confirms the official release and open-sourcing of the MiMo-V2.6 series on the launch date, describing it as "2 native fully multimodal models, namely Pro and Flash" with shared API pricing inherited from the V2.5 series. The post explicitly names Flash as a first-class member The official announcement states that MiMo-V2.6-Pro achieved a 46 score on the Artificial Analysis Intelligence Index, surpassing Kimi K3 and Qwen3.8 Max to become the most powerful open-source model at release, while flagging a remaining gap to closed-source leaders Claude Fable 5.1 and GPT-6 Astra — context that fram

Xiaomi Token Plan (Europe)

CoverageBenchmark

An in-depth third-party technical breakdown from eesel AI confirms that MiMo V2.6 is a family of open-weight, natively omnimodal language models from Xiaomi's MiMo team, officially released on September 22, 2026, with every checkpoint shipping a 1 million-token context window. The article explicitly distinguishes Flash The eesel AI analysis frames MiMo-V2.6-Flash within the broader open-model race, emphasizing that Xiaomi's release tagline is "Scaling Reinforcement Learning Toward Self-Improvement" — scaling RL compute on verifiable, complex tasks — rather than marketing-only positioning. It also walks through the official benchmarks

Xiaomi Token Plan (Europe)

Coverage

VentureBeat's September 22, 2026 launch coverage explicitly names MiMo-V2.6-Flash and reports its API pricing at $0.14 per million uncached input tokens and $0.28 per million output tokens, characterizing it as the second cheapest major frontier model available over API globally. The article confirms Flash retains the Alongside the Flash pricing detail, VentureBeat situates the variant within Xiaomi's broader MiMo-V2.6 launch: MIT-licensed open weights on Hugging Face, a top open-weight Intelligence Index score of 46 attributed to MiMo-V2.6-Pro (tied with Grok 4.7 and ahead of Grok 4.6, Gemini 3.8 Flash, DeepSeek V4.1 Flash and Pro)