Sulat.com
AI models
$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3 from Synthetic
Vercel AI Gateway logo

Model details

MiMo V2.5 Pro

MiMo V2.5 Pro is Xiaomi's flagship open-source model, designed for the most demanding agentic, complex software engineering, and long-horizon workloads. Xiaomi describes it as their most capable release to date, built to sustain complex trajectories across thousands of tool calls while maintaining strong instruction following and coherence over a very large context window. The model is released as open weights, making it accessible for teams that need a self-hostable foundation model without proprietary licensing constraints.

Under the hood, the model uses a Mixture-of-Experts design with 1.02 trillion total parameters and 42 billion active parameters, paired with a hybrid attention architecture that interleaves sliding window and global attention at a roughly 6:1 ratio to keep key-value cache storage compact without sacrificing long-context performance. A three-layer Multi-Token Prediction module, inherited from the earlier MiMo-V2-Flash lineage, helps speed up inference and rollout. Independent reporting notes that the model achieved an Artificial Analysis intelligence index score of 54 and successfully completed an autonomous Peking University compiler-development task over 4.3 hours and 672 tool calls, signaling practical strength for extended, tool-driven coding workflows.

Vercel AI Gatewayxiaomi/mimo-v2.5-promimo

Quick Info

Powered by
Provider
Vercel AI Gateway
Model key
xiaomi/mimo-v2.5-pro
Release date
Apr 22, 2026
Last updated
Apr 22, 2026
Knowledge cutoff
2024-12
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$0.435
Output token cost
$0.87

Limits

Output tokens
131,000 tokens
Context window
1,050,000 tokens

Transparent token rates

Compare MiMo V2.5 Pro 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 Pro

Vercel AI Gateway

Coverage

Xiaomi unveiled MiMo V2.5 and MiMo V2.5 Pro as open models, with the Pro tier positioned as surpassing Gemini 3.1 Pro and approaching Claude Opus 4.6 in performance. MiMo V2.5 Pro is a Mixture-of-Experts model with 1.02 trillion total parameters and 42 billion active parameters, versus the standard V2.5's 310 billion t In an autonomous compiler-development task based on a Peking University course, MiMo V2.5 Pro worked for 4.3 hours and completed a fully functional compiler after 672 tool calls, demonstrating long-horizon agentic capability. Artificial Analysis gave it an intelligence score of 54, surpassing DeepSeek V4 and Meta's Mus

Vercel AI Gateway

Coverage

The official XiaomiMiMo Hugging Face model card for MiMo-V2.5-Pro details an open-source Mixture-of-Experts model with 1.02T total and 42B active parameters, using a hybrid attention architecture that interleaves Sliding Window Attention and Global Attention in a 6:1 ratio with a 128 sliding window, paired with three l Post-training for MiMo-V2.5-Pro combines supervised fine-tuning, large-scale agentic reinforcement learning, and Multi-Teacher On-Policy Distillation (MOPD), targeting demanding agentic, complex software engineering, and long-horizon tasks with sustained trajectories over thousands of tool calls. The card also publishe

Videos about MiMo V2.5 Pro

More models around MiMo V2.5 Pro