DEV.co
Open-Source LLM · XiaomiMiMo

MiMo-V2-Flash

MiMo-V2-Flash is a 309B-parameter mixture-of-experts language model from Xiaomi with only 15B active parameters, designed for fast inference and agentic reasoning tasks. It uses a hybrid attention architecture and multi-token prediction to reduce computational overhead while maintaining competitive performance on standard benchmarks. The model is MIT-licensed, ungated, and available on HuggingFace.

Source: HuggingFace — huggingface.co/XiaomiMiMo/MiMo-V2-Flash
309.8B
Parameters
mit
License (OSI-approved)
Unknown
Context (tokens)
66.1k
Downloads (30d)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
DeveloperXiaomiMiMo
Parameters309.8B
Context windowUnknown
Licensemit — OSI-approved
Modality / tasktext-generation
Gated on HuggingFaceNo
Downloads66.1k
Likes743
Last updated2026-04-20
SourceXiaomiMiMo/MiMo-V2-Flash

What MiMo-V2-Flash is

MiMo-V2-Flash is a MoE architecture with 309B total parameters and 15B active parameters, trained on 27T tokens in FP8 mixed precision. It features: (1) hybrid attention interleaving sliding-window (128-token) and global attention at 5:1 ratio with learnable attention sink bias, reducing KV-cache by ~6x; (2) lightweight multi-token prediction module (0.33B params/block) for 3x faster inference; (3) native 32k context with 256k capability; (4) post-training via Multi-Teacher On-Policy Distillation and large-scale agentic RL. Evaluation shows competitive or superior performance on standard benchmarks (MMLU, GSM8K, code) versus larger models like DeepSeek-V3 and Kimi-K2, with particular strength on AIME, SWE-Bench, and long-context tasks.

Quickstart

Run MiMo-V2-Flash locally

Load the open weights with 🤗 Transformers and generate — the same model, self-hosted.

quickstart.pypython
from transformers import pipelinepipe = pipeline("text-generation", model="XiaomiMiMo/MiMo-V2-Flash")out = pipe("Explain retrieval-augmented generation in one sentence.",           max_new_tokens=128)print(out[0]["generated_text"])

Swap in vLLM or Ollama for production-grade serving. DEV.co can stand up the inference stack.

Deployment

How you'd run it

A typical self-hosted path — open weights, an inference server, your application.

DEV.co builds each layer — from GPU infrastructure to the application.

Best use cases

High-speed inference for agentic systems

With only 15B active parameters despite 309B total, this model excels in agent loops requiring rapid reasoning iterations. The multi-token prediction module triples output speed, reducing latency for tool-calling and planning workflows.

Software engineering and code-generation tasks

Strong performance on SWE-Bench (30.8%), AIME (35.3%), and code benchmarks (HumanEval+ 70.7%, MBPP+ 71.4%) indicates suitability for autonomous code review, bug fixing, and complex algorithmic problem-solving.

Long-context document analysis and RAG

Native 32k context with 256k capability and 99%+ accuracy on NIAH-Multi through 256k tokens makes this practical for processing lengthy documents, legal reviews, and retrieval-augmented generation without chunking overhead.

Running & fine-tuning it

ESTIMATE: ~30–40 GB VRAM for FP8 inference (15B active params × 2 bytes + KV-cache + overhead). Full precision inference would require 60+ GB. Multi-GPU setup recommended for production (<100ms latency). Requires recent NVIDIA H100/A100 or equivalent for efficient serving. Exact requirements depend on batch size, context length, and serving framework; vendor documentation should be reviewed.

Not explicitly stated in card. MoE architectures typically support LoRA on expert FFNs and query/key projections, but custom-code requirement suggests non-standard layer structure. Feasibility of QLoRA or parameter-efficient fine-tuning Unknown without architecture details. Recommend: (1) consult GitHub repository for training code, (2) test LoRA on base model before production tuning, (3) assess activation sparsity impact on LoRA gradient flow.

When to avoid it — and what to weigh

  • Real-time edge inference on resource-constrained devices — Despite efficiency gains, 15B active parameters still requires substantial VRAM (estimated 30–40 GB in FP8). Not suitable for mobile, IoT, or single-GPU edge deployment without quantization beyond FP8.
  • Production use without custom integration testing — Model card lists 'custom_code' tag, indicating non-standard architecture components. Deployment via standard frameworks (vLLM, TGI) requires verification; vendor-specific tooling may be necessary.
  • Use cases requiring strict Chinese language performance — On Chinese benchmarks (C-Eval, CMMLU, C-SimpleQA), performance lags Kimi-K2 and DeepSeek by 3–16 percentage points. English and code tasks are stronger relative to scale.
  • Scenarios with unknown hardware availability — Serves as a large model that requires enterprise-grade infrastructure (multi-GPU, high-bandwidth interconnect) for production. Smaller quantized alternatives may be preferable for constrained environments.

License & commercial use

MIT license. This is a permissive OSI-approved license that permits commercial use, modification, and distribution provided the license and copyright notice are retained.

MIT license explicitly permits commercial use without restriction. However, deployment of 309B-parameter model incurs significant infrastructure costs (GPUs, bandwidth). Verify with legal team: (1) liability indemnification in vendor terms of service; (2) model weight distribution compliance if bundling into proprietary product; (3) absence of export restrictions on Xiaomi-developed artifacts. No gatekeeping or usage restrictions observed on HuggingFace, but review Xiaomi MiMo API Platform's commercial terms if using managed hosting.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationStrong
License clarityClear
Deployment complexityModerate
DEV.co fitStrong
Assessment confidenceHigh
Security considerations

Not stated in card. General considerations: (1) model weights downloaded from HuggingFace should be verified against official SHA/PGP signatures if available (check GitHub); (2) 'custom_code' tag requires code review before execution in production—audit any imports and dynamic behavior; (3) no mention of adversarial robustness testing or jailbreak evaluation; (4) long-context capability (256k tokens) may increase prompt-injection surface area; (5) review Xiaomi privacy policy if using managed API Platform. Do not assume the model is adversarially robust or free of biases without independent evaluation.

Alternatives to consider

DeepSeek-V3 (Base or Exp variants)

Similar scale (37B active/671B total) with stronger Chinese performance and comparable math/code scores. More established ecosystem and broader deployment support. Trade-off: less aggressive efficiency gains on English benchmarks.

Kimi-K2

32B active/1043B total with best-in-class Chinese and conversational capabilities. Stronger on HellaSwag, SimpleQA, TriviaQA. Trade-off: higher parameter count, no public release roadmap, potential licensing constraints from Zhipu AI.

Llama 3.1 405B (or quantized variants)

Fully open-source, well-established fine-tuning ecosystem, native support across all major inference frameworks. Trade-off: significantly larger (405B), no MoE efficiency gains, lower performance on some benchmarks like AIME but stronger on code.

Software development agency

Ship MiMo-V2-Flash with senior software developers

Download the model from HuggingFace, test on your target hardware, and review the technical report. Consult the GitHub repository for architecture details and fine-tuning guidance. Contact Devco for infrastructure sizing and custom integration support.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

MiMo-V2-Flash FAQ

Can I use MiMo-V2-Flash in a commercial product?
Yes. The MIT license explicitly permits commercial use. However, you must: (1) retain the license and copyright notice in distributions, (2) verify compliance with Xiaomi's terms of service if using their managed API, (3) assume liability for model outputs (no indemnification stated), and (4) consult legal counsel on export restrictions and data privacy for your specific jurisdiction and use case.
What GPU do I need to run this model?
Estimate 30–40 GB VRAM for FP8 inference at batch size 1 (15B active params + KV-cache). For production (multiple concurrent requests), plan for 2–4 high-end GPUs (H100, A100) or equivalent. Exact requirements depend on your serving framework, batch size, and context length. Test on your target hardware before deployment.
How does MiMo-V2-Flash compare to GPT-4 or Claude?
Direct comparison not provided in card. MiMo-V2-Flash shows strong performance on standard academic benchmarks (MMLU 86.7%, GSM8K 92.3%, SWE-Bench 30.8%) but is open-source and not directly compared to proprietary models. Capabilities differ: no multi-modal input, no real-time web search, training data and cutoff date Unknown. Evaluate on your specific use case.
Does it support fine-tuning or LoRA?
Not explicitly documented. The 'custom_code' tag suggests non-standard architecture (MoE routing, multi-token prediction). LoRA feasibility is Unknown without detailed layer specification. Recommendation: test LoRA on the base model first, or consult the GitHub repository and Xiaomi support for guidance on parameter-efficient tuning strategies.

Work with a software development agency

Need help beyond evaluating MiMo-V2-Flash? DEV.co is a software development agency offering software development services and web development for teams of every size. Our software developers and web developers build custom software, web applications, APIs, and open-source llms integrations — and maintain them long-term.

Evaluate MiMo-V2-Flash for Your Use Case

Download the model from HuggingFace, test on your target hardware, and review the technical report. Consult the GitHub repository for architecture details and fine-tuning guidance. Contact Devco for infrastructure sizing and custom integration support.