Xiaomi model product
MiMo V2.5 Pro
02T-parameter sparse Mixture-of-Experts language model with 42B active parameters and a 1M-token context window.
Updated Aug 12, 2026. Default version: MiMo-V2.5-Pro
LLMBoard score46.7MiMo-V2.5-Pro
Coverage80%27 benchmark families
Context window1MTokens
Official input price$0.435Xiaomi API
MiMo V2.5 Pro Specifications
Technical details for the model's default version.
- Version
- MiMo-V2.5-Pro
- Released
- Apr 27, 2026
- Knowledge cutoff
- Dec 1, 2024
- Parameters
- 1T
- Context window
- 1M
- Max output
- 131.1K
- Inputs
- text
- Outputs
- text
- Open weights
- Yes
- License
- MIT
MiMo V2.5 Pro Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
MiMo V2.5 Pro Benchmark Results
Benchmark scores for MiMo-V2.5-Pro.
| ARC-C | 97.2% | 01 | 34 | 100.0% | C | |
| CMMLU | 90.2% | 01 | 6 | 100.0% | C | |
| FrontierSWE (Impl.) | 3.4 rank | 01 | 1 | 100.0% | C | |
| Global-MMLU | 83.6% | 01 | 5 | 100.0% | C | |
| GSM8k | 99.6% | 01 | 48 | 100.0% | C | |
| MBPP+ | 74.1% | 01 | 4 | 100.0% | C | |
| MiMo Coding Bench | 73.7% | 01 | 2 | 100.0% | C | |
| TAU3-Bench | 72.9% | 01 | 5 | 100.0% | C | |
| AIME | 37.3% | 02 | 2 | 0.0% | C | |
| BBH | 88.4% | 02 | 12 | 90.9% | C | |
| SWE-bench Verified (Agentless) | 35.7% | 02 | 2 | 0.0% | C | |
| Winogrande | 85.6% | 02 | 22 | 95.2% | C | |
| GraphWalks | 62.0% | 03 | 3 | 0.0% | C | |
| TriviaQA | 81.3% | 03 | 18 | 88.2% | C | |
| DROP | 86.3% | 04 | 30 | 89.7% | C | |
| HellaSwag | 89.8% | 04 | 27 | 88.5% | C | |
| C-Eval | 91.5% | 05 | 18 | 76.5% | C | |
| WildClawBench | 43.0% | 05 | 5 | 0.0% | C | |
| Claw-Eval | 64.0% | 06 | 13 | 58.3% | C | |
| HumanEval+ | 75.6% | 07 | 10 | 33.3% | C | |
| MATH | 86.2% | 11 | 71 | 85.7% | C | |
| Terminal-Bench 2.0 | 68.4% | 13 | 49 | 75.0% | C | |
| Finance Agent v2 | 41.5% | 15 | 26 | 44.0% | B | |
| MMLU | 89.4% | 15 | 100 | 85.9% | C | |
| MMLU-Redux | 92.8% | 15 | 48 | 70.2% | C | |
| SWE-Bench Verified | 78.9% | 17 | 105 | 84.6% | C | |
| SWE-Bench Pro | 57.2% | 23 | 45 | 50.0% | C | |
| Humanity's Last Exam | 34.0% | 42 | 93 | 55.4% | C | |
| LiveCodeBench v6 | 39.6% | 50 | 53 | 5.8% | C | |
| MMLU-Pro | 68.5% | 90 | 129 | 30.5% | C | |
| GPQA | 66.7% | 141 | 234 | 39.9% | C |
MiMo V2.5 Pro Arena Results
Preference and agent-evaluation results for the default version.
| text | overall | 25 | 1464.7 | 50,109 | N/A | |
| text factuality | overall | 28 | 1465.0 | 50,023 | N/A | |
| agent bash recovery steps | overall | 30 | 0.0 | N/A | 31.7K | |
| agent steerability | overall | 30 | -0.0 | N/A | 27.7K | |
| agent | overall | 32 | -0.0 | N/A | 1.1M | |
| agent praise complaint | overall | 34 | -0.1 | N/A | 6.5K | |
| agent task outcome explicit | overall | 35 | -0.0 | N/A | 17.2K | |
| webdev | overall | 35 | 1474.5 | 12,927 | N/A | |
| text style control | overall | 37 | 1467.5 | 50,109 | N/A | |
| agent tool hallucination | overall | 38 | -0.0 | N/A | 1M |
MiMo V2.5 Pro Runtime Performance
Provider-specific output speed and catalog latency for MiMo-V2.5-Pro. Runtime does not affect the capability score.
| DeepInfra | 195.755 tok/s | 0.625 s | 1M | 131.1K | |
| Xiaomi | 135.012 tok/s | 2.61 s | 1M | 131.1K | |
| Novita | 130.693 tok/s | 2.845 s | 1M | 131.1K |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
MiMo V2.5 Pro Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- $0.435 input, $0.87 output per 1M
- Official provider
- Xiaomi
- Lowest third-party
- From $0.40 input, $0.80 output per 1M via CrofAI
- Tracked offerings
- 22
| Kenari | mimo-v2-5-pro | global | N/A | N/A | 1M | |
| Xiaomi Token Plan (China) | mimo-v2.5-pro | global | N/A | N/A | 1M | |
| Xiaomi Token Plan (Europe) | mimo-v2.5-pro | global | N/A | N/A | 1M | |
| Xiaomi Token Plan (Singapore) | mimo-v2.5-pro | global | N/A | N/A | 1M | |
| CrofAI | mimo-v2.5-pro | global | $0.40 | $0.80 | 1M | |
| DigitalOcean | mimo-v2.5-pro | global | $0.40 | $1.5 | 262.1K | |
| Kilo Gateway | xiaomi/mimo-v2.5-pro | global | $0.40 | $1.5 | 1M | |
| OpenCode Go | mimo-v2.5-pro | global | $0.435 | $0.87 | 1M | |
| NanoGPT | xiaomi/mimo-v2.5-pro | global | $0.435 | $0.87 | 1M | |
| LLM Gateway | mimo-v2.5-pro | global | $0.435 | $0.87 | 1M | |
| Pioneer | XiaomiMiMo/MiMo-V2.5-Pro | global | $0.435 | $0.87 | 1.1M | |
| Vercel AI Gateway | xiaomi/mimo-v2.5-pro | global | $0.435 | $0.87 | 1.1M | |
| OpenRouter | xiaomi/mimo-v2.5-pro | global | $0.435 | $0.87 | 1.1M | |
| Xiaomi | mimo-v2.5-pro | global | $0.435 | $0.87 | 1M | |
| CrossModel | xiaomi/mimo-v2.5-pro | global | $0.47 | $0.94 | 1M | |
| NovitaAI | xiaomimimo/mimo-v2.5-pro | global | $0.522 | $1.04 | 1M | |
| Vultr | XiaomiMiMo/MiMo-V2.5-Pro | global | $0.55 | $1.65 | 1M | |
| ZenMux | xiaomi/mimo-v2.5-pro | global | $1 | $3 | 1M | |
| Hugging Face | XiaomiMiMo/MiMo-V2.5-Pro | global | $1 | $3 | 1M | |
| Deep Infra | XiaomiMiMo/MiMo-V2.5-Pro | global | $1 | $3 | 1M | |
| ClinePass | cline-pass/mimo-v2.5-pro | global | $1.74 | $3.48 | 1M | |
| EmpirioLabs AI | mimo-v2-5-pro | global | $2.18 | $4.35 | 1M |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
MiMo V2.5 Pro Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| MiMo-V2.5-Pro | 46.7 | 1T | 1M | 131.1K | Yes | MIT |
What is MiMo V2.5 Pro?
Key information about MiMo V2.5 Pro and its available data.
02T-parameter sparse Mixture-of-Experts language model with 42B active parameters and a 1M-token context window.
It inherits the MiMo-V2-Flash hybrid-attention and Multi-Token Prediction design, extends context during pre-training up to 1M tokens, and uses supervised fine-tuning, domain-specialized reinforcement learning, and Multi-Teacher On-Policy Distillation to improve complex software engineering, long-horizon agentic tasks, and ultra-long-context coherence.
Data as of 2026-08-11.
MiMo V2.5 Pro vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to MiMo V2.5 Pro
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
XI
MiMo V2 Flash
Xiaomi
XI
MiMo V2 Omni
Xiaomi
XI
MiMo
Xiaomi
XI
MiMo V2 Pro
Xiaomi
XA
Grok 4.1
xAI
OP
GPT-OSS-20B
OpenAI
FAQ
Common questions about MiMo V2.5 Pro.
When was MiMo V2.5 Pro released?
MiMo V2.5 Pro's default version was released on Apr 27, 2026.
How much does MiMo V2.5 Pro cost?
MiMo V2.5 Pro's official API price is $0.435 per million input tokens and $0.87 per million output tokens via Xiaomi. The lowest tracked third-party offer starts at $0.40 input and $0.80 output via CrofAI.
Who created MiMo V2.5 Pro?
MiMo V2.5 Pro was created by Xiaomi.
What is the context window for MiMo V2.5 Pro?
The default version has a 1M token context window.
Is MiMo V2.5 Pro open weight?
Yes. The default version is marked as open weight under MIT.
How many API providers offer MiMo V2.5 Pro?
22 provider offerings are linked to the default version.
What models should I compare MiMo V2.5 Pro with?
Nearby ranked alternatives include GPT-5.4-nano, Qwen3 235B A22B Thinking, Grok 3 Mini.