MiMo-V2.6-Flash

MiMo-V2.6-Flash sits inside Xiaomi's MiMo-V2.6 series as the lower-cost counterpart to the flagship MiMo-V2.6-Pro, built to balance intelligence, efficiency, and speed for latency-sensitive professional work. The series is described as natively omnimodal, handling text along with visual inputs and oriented toward coding tasks, computer use, and complex reasoning across multi-agent collaborations. Xiaomi's MiMo platform frames the lineup as full-modality and engineered for the Pareto frontier of intelligence, cost, and speed, with Flash specifically positioned as the efficiency-oriented tier within that design philosophy.

According to external launch reporting, MiMo-V2.6-Flash is configured with roughly 309 billion total parameters and about 15 billion active parameters, a sparse-activation design intended to keep inference economical while preserving capability for large-scale coding and reasoning workloads. The series launch was paired with an unusual degree of process transparency, including a public livestream of the reinforcement-learning training run that preceded the release. Flash also benefits from the family's Pro-UltraSpeed option, which reportedly delivers up to twenty times faster output at equivalent quality, making it well suited to latency-sensitive pipelines that still need strong multimodal understanding.

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

Open provider page Provider docs

Quick Info

Provider: Xiaomi Token Plan (Singapore)
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 (Singapore)

Coverage

According to VentureBeat's September 21, 2026 coverage of Xiaomi's MiMo-V2.6 release, MiMo-V2.6-Flash is positioned as a smaller, substantially cheaper sibling to the flagship MiMo-V2.6-Pro, priced at $0.14 per million uncached input tokens and $0.28 per million output tokens via Xiaomi's API. The article states Flash The same VentureBeat report frames the release in the context of Xiaomi's broader V2.6 launch, noting both models are open weight and MIT-licensed, downloadable from Hugging Face, and accessible via Xiaomi's API. While the headline benchmark figures (Artificial Analysis Intelligence Index of 46, output speed ~134 token

Xiaomi Token Plan (Singapore)

CoverageBenchmark

The llm-stats.com benchmark dashboard for MiMo-V2.6-Flash reports a blended price of $0.15 per million tokens alongside an LLM Stats Score of 45.4, placing it among the top cost-efficient frontier-tier models tracked on the site. The page sources performance data from mimo.xiaomi.com and lists benchmark rankings includ The dashboard contextualizes Flash within its leaderboard alongside Gemma 4 E4B, GPT OSS 120B, DeepSeek-V4-Flash-0731, DeepSeek-V4.1-Flash, and Claude Opus 5.5, comparing cost efficiency and LLM Stats Score at blended prices plotted on a log scale. Conversation-depth performance tracking for Flash shows Turn 1 scores a

Xiaomi Token Plan (Singapore)

Official sourceDocumentation

The English-language official Xiaomi MiMo release page announces the open-sourcing of the MiMo-V2.6 series, comprising two native fully multimodal models: Pro and Flash. It confirms the V2.6 series adopts the same API pricing as V2.5, keeping the cost structure unchanged while intelligence is upgraded. MiMo-V2.6-Pro is The official English post frames V2.6 as a step along Xiaomi's recursive self-improvement path, scaling reinforcement-learning compute on verifiable complex tasks. It states that during the RL training phase, MiMo-V2.6 is likely one of the domestic open-source models allocated the largest amount of compute to date, wit