MiMo-V2.6-Flash
Overview
MiMo-V2.6-Flash is an open-weight, native multimodal reasoning model released by Xiaomi MiMo on September 22, 2026. Available through the API as mimo-v2.6-flash, it combines a sparse Mixture-of-Experts architecture with a 1M-token context window, targeting frequent calls, large-scale office workloads, and cost-sensitive coding and agent applications.
Key Features
- Efficient Sparse Architecture: Uses 309 billion total parameters with 15 billion activated parameters; the published MiMo-V2.6-Flash-RL checkpoint is MIT-licensed.
- Native Multimodal Understanding: Processes text, images, video, and audio together, enabling workflows that combine written instructions with visual and acoustic evidence.
- Long-Context Agent Work: Supports extensive source material, repository context, and tool histories within a 1M-token window.
- Broad Agent Training: Mixed reinforcement learning spans coding, general agents, visual tasks, and cybersecurity across multiple agent frameworks.
- Flexible API Workflows: Supports switchable deep thinking, function calls, streaming, JSON output, prompt caching, and discounted asynchronous batch processing.
Best Use Cases
- High-Volume Office Automation: Summarization, classification, and structured extraction where lower token prices and batch processing help control aggregate cost.
- Visual Workflow Agents: Interpreting screenshots and interface states, then using external tools to complete and check routine computer tasks.
- Iterative Coding Assistance: Implementation, debugging, and test-and-revise loops that combine code, tool results, and visual feedback.
- Multimodal Document Analysis: Reviewing long collections of text, charts, recordings, and video with shared context.
Capabilities and Limitations
| Capability | Description |
|---|---|
| Reasoning | Deep thinking is enabled by default and can be disabled through thinking.type in Chat Completions. |
| Coding | Xiaomi's released model-card evaluation reports 67.9 on DeepSWE v1.1. |
| Agent Evaluation | Xiaomi reports 73.6 on Toolathlon-Verified and 80.8 on OSWorld-Verified. These are provider-reported benchmark results, not production guarantees. |
| Multimodal | Text, image, video, and audio input; text output. |
| Context Window | 1M tokens. |
| Max Output | 131,072 tokens, shared by reasoning and the final response. |
| Tool Use | Function calling and Xiaomi's separately billed web-search service. |
| Structured Output | JSON object mode; fields, types, and nesting remain prompt-defined and require application-side validation. |
Known Limitations
- Native output is text. Computer operation and media creation require external tools or generated code.
- Thinking mode forces
temperature=1.0andtop_p=0.95. Multi-turn tool conversations must preserve historicalreasoning_content; missing it can cause HTTP 400 errors. - Chat Completions supports only
tool_choice=auto; other values are discarded and treated as automatic selection.
Standard Pricing
The token prices below are shown in USD per 1 million tokens; web search is billed in USD per use.
| Model | Input (USD / 1M Tokens) | Cache Write (USD / 1M Tokens) | Cache Read (USD / 1M Tokens) | Output (USD / 1M Tokens) | Web Search (USD / use) |
|---|---|---|---|---|---|
| MiMo-V2.6-Flash | $0.14 | $0.14 | $0.0028 | $0.28 | - |
Credits settlement: B.AI converts charges at 1 USD = 1,000,000 Credits and deducts Credits from the account balance.
Prices shown in the documentation are B.AI standard reference prices for base billing purposes. B.AI may provide lower actual usage costs through top-up bonuses and account benefits. Specific prices, bonus Credits, and account benefits are subject to the platform display and final billing records.