MiMo-V2.5

MiMo-V2.5-Pro represents Xiaomi's flagship entry in the MiMo V2.5 lineup, replacing the earlier MiMo-V2 series, which was officially deprecated on June 30 to push users toward the newer generation. Built around a trillion-scale sparse architecture with roughly 1T total parameters and 42B active at inference, the model is engineered to handle extremely long horizons, including autonomous agent tasks that involve more than a thousand tool calls. Open weights are published under the XiaomiMiMo organization on Hugging Face, reflecting Xiaomi's commitment to letting developers self-host and fine-tune the model rather than relying solely on a closed API.

The flagship variant targets demanding agentic and software-engineering workloads, posting top placements on benchmarks such as ClawEval, GDPVal, and SWE-bench Pro, where Xiaomi positions it as rivalling top frontier agent systems on complex, multi-day professional tasks. Its 1M-token context window makes it a natural fit for integration with agent frameworks and coding assistants that need to reason over large codebases, long documents, or extended tool traces. Community experimentation has shown the model runnable across a pair of DGX Spark nodes, underscoring that the sparse activation pattern keeps the effective footprint manageable despite the trillion-parameter total scale, which broadens practical deployment options for teams with limited hardware budgets.

Xiaomimimo-v2.5mimodeprecated

Open provider page Provider docs

Quick Info

Provider: Xiaomi
Model key: mimo-v2.5
Release date: Apr 22, 2026
Last updated: Apr 22, 2026
Knowledge cutoff: 2024-12
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.14
Output token cost: $0.28

Limits

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

Transparent token rates

Compare MiMo-V2.5 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-V2.5

Xiaomi

Official sourceDocumentation

One-time purchase unlocks both MiMo-V2.5 flagship models, plus TTS model free across all tiers for a limited time. Unleash powerful productivity with Xiaomi MiMo

OpenRouter

CoverageBenchmark

A community-authored NVIDIA developer-forums thread (a3refaat, May 2026) documents a reproducible recipe for serving a third-party NVFP4 quantization of MiMo-V2.5 (lukealonso/MiMo-V2.5-NVFP4) on a 2× DGX Spark / GB10 cluster. The setup uses vLLM 0.21.1rc1.dev39 on CUDA 13.2 with tensor parallelism = 2, instanttensor lo Three concrete correctness fixes are detailed: (1) the ModelOpt mixed-precision path must dispatch MXFP8 linear layers to the MXFP8 method rather than falling through when the checkpoint mixes MXFP8 dense layers with NVFP4 experts; (2) the weight_scale_inv field is UE8M0 MXFP8 scale metadata and should not be reciproca

Xiaomi

Coverage

Xiaomi has announced a major upgrade to its AI voice ecosystem with the launch of the MiMo-V2.5-TTS series and MiMo-V2.5-ASR

OpenRouter

CoverageBenchmark

Xiaomi's MiMo-V2.5 is described on OpenRouter as a native omnimodal model handling image and video understanding, with a listed context window of 1.1M tokens suited to whole documents, long conversations, and agent-framework integrations. The page names it explicitly as xiaomi/mimo-v2.5 with model weights and API access, and states it delivers Pro-level agentic performance at roughly half the inference cost of prior tiers. Benchmark figures on the listing span GPQA Diamond scores of 81.3% to 82.8% depending on endpoint, with TAU-Bench at 79.0% to 79.3% and a StreamLake column also reported. Xiaomi is credited as the creator of the model, and the listing records a release date of April 22, 2026, though the capability comparisons and scores are aggregator-supplied without published methodology.

Sulat also reports on AI and tech beyond the catalog.Read news.sulat.com