Model reference · open weights
MiMo-Pro is an open-weight language model from XiaomiMiMo. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.
Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.
What it is
| Released by | XiaomiMiMo |
|---|---|
| Type | Language models |
| Task | Text gen |
| Parameters (lead) | 1023.2B |
| Context | 1024k tokens |
| Runs with | transformers |
| Released | 2026-04-27 |
| Popularity | 25k downloads / month |
| Licence | Open weights |
About
| | | | |
| | |
MiMo-V2.5-Pro is an open-source Mixture-of-Experts (MoE) language model with 1.02T total parameters and 42B active parameters. It utilizes the hybrid attention architecture and 3-layers Multi-Token Prediction (MTP) introduced in MiMo-V2-Flash, with up to 1M tokens context length.
MiMo-V2.5-Pro is our most capable model to date, designed for the most demanding agentic, complex software engineering, and long-horizon tasks. It sustains complex trajectories spanning thousands of tool calls with strong instruction following and coherence over a 1M-token context window. Key features include:
| Model | Total Params | Active Params | Context Length | Precision | Download |
|---|---|---|---|---|---|
| MiMo-V2.5-Pro | 1.02T | 42B | 1M | FP8 (E4M3) Mixed | 🤗 HuggingFace 🤖 ModelScope |
| MiMo-V2.5-Pro-Base | 1.02T | 42B | 256K | FP8 (E4M3) Mixed | 🤗 HuggingFace 🤖 ModelScope |
| Category | Benchmark | Setting | MiMo-V2.5-Pro Base | MiMo-V2.5 Base | DeepSeek-V4-Pro Base | DeepSeek-V4-Flash Base | Kimi-K2 Base |
|---|---|---|---|---|---|---|---|
| Params | #Activated / #Total | - | 42B / 1.02T | 15B / 310B | 49B / 1.6T | 13B / 284B | 32B / 1.04T |
| General | BBH | 3-shot | 88.4 | 87.2 | 87.5 | 86.9 | 88.7 |
| MMLU | 5-shot | 89.4 | 86.3 | 90.1 | 88.7 | 87.8 | |
| MMLU-Redux | 5-shot | 92.8 | 89.8 | 90.8 | 89.4 | 90.2 | |
| MMLU-Pro | 5-shot | 68.5 | 65.8 | 73.5 | 68.3 | 69.2 | |
| DROP | 3-shot | 86.3 | 83.7 | 88.7 | 88.6 | 83.6 | |
| ARC-Challenge | 25-shot | 97.2 | 96.5 | - | - | 96.2 | |
| HellaSwag | 10-shot | 89.8 | 88.6 | 88.0 | 85.7 | 94.6 | |
| WinoGrande | 5-shot | 85.6 | 84.7 | 81.5 | 79.5 | 85.3 | |
| TriviaQA | 5-shot | 81.3 | 80.7 | 85.6 | 82.8 | 85.1 | |
| GPQA-Diamond | 5-shot | 66.7 | 58.1 | - | - | 48.1 | |
| Math | GSM8K | 8-shot | 99.6 | 83.3 | 92.6 | 90.8 | 92.1 |
| MATH | 4-shot | 86.2 | 67.7 | 64.5 | 57.4 | 70.2 | |
| AIME 24&25 | 2-shot | 37.3 | 36.9 | - | - | 31.6 | |
| Code | HumanEval+ | 1-shot | 75.6 | 71.3 | - | - | 84.8 |
| MBPP+ | 3-shot | 74.1 | 70.9 | - | - | 73.8 | |
| LiveCodeBench v6 | 1-shot | 39.6 | 35.5 | - | - | 26.3 | |
| SWE-Bench (AgentLess) | 3-shot | 35.7 | 30.8 | - | - | 28.2 | |
| Chinese | C-Eval | 5-shot | 91.5 | 88.6 | 93.1 | 92.1 | 92.5 |
| CMMLU | 5-shot | 90.2 | 88.2 | 90.8 | 90.4 | 90.9 | |
| Multilingual | GlobalMMLU | 5-shot | 83.6 | 77.4 | - | - | 80.7 |
GraphWalks is a long-context benchmark from OpenAI that fills the prompt with a directed graph of hex-hash nodes and asks the model to run a breadth-first search (nodes exactly at depth N) or list a node's parents. We evaluate across the full 32k–1M input-token span and apply the same evaluation fixes described by Anthropic.
MiMo V2.5 Pro delivers a major leap in long-context reasoning. Past 128k, V2 Pro degrades rapidly and collapses to 0.00 at 1M on both subtasks, while V2.5 Pro still scores 0.56 BFS / 0.92 Parents at 512k and 0.37 / 0.62 at 1M.
MiMo-V2.5-Pro addresses the quadratic complexity of long contexts by interleaving Local Sliding Window Attention (SWA) and Global Attention (GA). Unlike traditional speculative decoding, our MTP module is natively integrated for training and inference.
| Component | MiMo-V2.5-Pro | MiMo-V2.5 |
|---|---|---|
| Total Parameters | 1.02T | 310B |
| Activated Parameters | 42B | 15B |
| Hidden Size | 6144 | 4096 |
| Num Layers | 70 (1 dense + 69 MoE) | 48 (1 dense + 47 MoE) |
| Full Attention Layers | 10 | 9 |
| SWA Layers | 60 | 39 |
| Num Attention Heads | 128 | 64 |
| Num KV Heads | 8 (GQA) | 8 (GA) / 4 (SWA) |
| Head Dim (QK / V) | 192 / 128 | 192 / 128 |
| Routed Experts | 384 | 256 |
| Experts per Token | 8 | 8 |
| MoE Intermediate Size | 2048 | 2048 |
| Dense Intermediate Size | 16384 (layer 0 only) | 16384 (layer 0 only) |
| SWA Window Size | 128 | 128 |
| Max Context Length | 1M | 1M |
| MTP Layers | 3 | 3 |
For post-training, MiMo-V2.5-Pro adopts the three-stage post-training paradigm introduced in MiMo-V2-Flash to achieve exceptional performance. The paradigm begins with Supervised Fine-Tuning (SFT) to build strong, foundational instruction-following skills using curated data pairs. Next, in the Domain-Specialized Training stage, diverse teacher models — ranging from math and safety to complex agentic tool-use — are individually optimized using domain-specific RL rewards. Finally, the process culminates in Mult
From the published model card. Full card on the HuggingFace links in the sidebar.
Using it via the API
Once AxForge deploys mimo-pro for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (mimo-pro below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/chat/completions \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"mimo-pro","messages":[{"role":"user","content":"Hello"}]}'
Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.