Xiaomi MiMo-V2.6-Pro (1T MoE) Architecture Parameters, Benchmarks & VRAM
Xiaomi MIT-licensed 1T-parameter omni-modal reasoning flagship with 1.05M context.
🧬 Xiaomi MiMo-V2.6-Pro (1T MoE) Architecture & Tensor Specifications
Verified directly against HuggingFace (XiaomiMiMo/MiMo-V2.6-Pro-RL) config.json & Model Card
1. Backbone & Dimensions
mimo_v2- Total / Active Params
- 1.024T / 28B
- Hidden Size (d_model)
- 6,144
- FFN Intermediate Size
- 16,384
- MoE Expert Dim
- 2,048
- MoE Experts
- 384 Routed · Top-8
- Activation / Norm
- SiLU (SwiGLU) · ε=1e-6
- Native Tensor Dtype
- BF16 / FP8
2. Attention & Layer Layout
GQA- Num Hidden Layers
- 70 Layers
- Hidden Layer Layout
- 70-Layer Sparse MoE (384 Experts, Top-8 Active)
- Attention Heads (Q / KV)
- 128 Q / 8 KV (16:1 GQA)
- Head Dimension
- 192
- Vision Encoder / MTP
- ViT 28L
3. Context, RoPE & Tokenizer
Tokenizer- Native Context Length
- 1,048,576 tokens
- RoPE Configuration
- RoPE (原生 1,048,576 Context)
- Vocab Size
- 152,576 (Untied)
📋 Xiaomi MiMo-V2.6-Pro (1T MoE) Official API & System Card Specifications
Verified against official Xiaomi 小米 MiMo developer documentation & pricing
1. Identity & Capabilities
API Spec- API Model ID
- mimo-v2.6-pro
- Release Date
- 2026-09-22
- Weight License
- Open-Weight (API Spec)
- Native Reasoning
- ✅ Supported
- Function Calling & JSON Schema
- ✅ 原生支持
- Batch API 异步批处理
- ✅ Supported (50% off)
2. Context, Output & Modalities
Limits- Max Context Window
- 1,024,000 tokens (1024K)
- Max Output Tokens
- 131,000 tokens
- Input Modalities
- Text + Image + Video
- Output Modalities
- Text
- Direct China Access
- ✅ Direct CN + RMB
- Deployment Mode
- Open-Weight + Cloud API
3. Official Pricing & Cache
CNY / 1M- Input / 1M Tokens
- ¥ 3
- Cached Input / 1M
- ¥ 0.02 -99%
- Output / 1M Tokens
- ¥ 6
- Tiered Pricing Notes
- Xiaomi MiMo: Xiaomi MiMo official API platform
🎛️ Official Sampling & Reasoning Parameters
Recommended generation settings & API parameter constraints
📊 Xiaomi MiMo-V2.6-Pro (1T MoE) Full Benchmark Scorecard
16 public benchmarks · ranks among 154 models · synced 2026-11-20
💻 Coding 9 items
CyberGym is a benchmark for evaluating AI agents on cybersecurity tasks, testing their ability to identify vulnerabilities, perform security analysis, and complete security-related challenges in a controlled environment.
Terminal-Bench 2.1 is an updated release of the Terminal-Bench benchmark that tests AI agents' ability to operate a computer via the terminal.
DeepSWE 1.1 evaluates software engineering agents using the mini-swe-agent harness.
SEC-bench Pro is a self-evolving software-security benchmark that measures agent bug hunting on critical, high-complexity systems.
MiMo Coding Bench evaluates coding-agent capabilities on software engineering tasks reported with the MiMo model family.
ExploitBench is a cybersecurity benchmark that evaluates a model's ability to discover and exploit software vulnerabilities, reported as the fraction of challenges where the model captures the target (Cap%).
Terminal-Bench 4.0 evaluates AI agents on real-world work in terminal and command-line containerized environments across 66 tasks, with emphasis on science-adjacent and frontier engineering problems.
Program Bench evaluates code-generation agents by asking them to recreate a program's behavior from only a compiled binary and documentation.
ExploitGym is a large-scale, realistic benchmark built from real-world vulnerabilities across userspace programs, Google's V8 engine, and the Linux kernel.
🤖 Agents & Tools 6 items
OSWorld-Verified is a verified subset of OSWorld, a scalable real computer environment for multimodal agents supporting task setup, execution-based evaluation, and interactive learning across Ubuntu, Windows, and macOS.
Verified Toolathlon evaluation of agent tool-use capability, as reported in the MiMo-V2.6 release.
JobBench evaluates agents on work-related tasks, as reported in the MiMo-V2.6 release.
Version 2.1 of the Artificial Analysis GDPval evaluation of professional knowledge work, reported as an Elo rating.
AutomationBench v1.0.6 evaluates agent performance on multi-step workflow automation tasks.
Agents' Last Exam is a challenging benchmark for AI agents on hard, long-horizon tasks that test sustained reasoning, planning, and tool use, reported with and without tool access.
Source: ZeroEval / llm-stats.com (✓ = third-party verified, others vendor-reported).
⚡ Xiaomi MiMo-V2.6-Pro (1T MoE) Interactive VRAM & Token Speed Calculator
Physical 96 GB · Load 681.9%
Exceeds usable VRAM by 547.48 GB
Enable Q8/Q4 KV quantization to scale batch concurrency
OLLAMA_CONTEXT_LENGTH=16384 ollama run xiaomi-mimo-v2-6-proInsufficient Local VRAM? Ways to Run Xiaomi MiMo-V2.6-Pro (1T MoE)
Rent an hourly high-VRAM cloud GPU pod, or check the 12-month Buy vs Cloud vs API break-even curve.
📊 Xiaomi MiMo-V2.6-Pro (1T MoE) Quantization VRAM Breakdown (16K Context)
| Quantization | Bits/Weight | Weights | KV Cache (16K) | Total VRAM | Quality Note |
|---|---|---|---|---|---|
| FP16 / BF16 | 16 bpw | 2048 GB | 7.05 GB | 2097.31 GB | Lossless Full Precision |
| FP8 / Q8_0 | 8.5 bpw | 1088 GB | 7.05 GB | 1118.11 GB | 99.9% Near-Lossless |
| GGUF Q6_K | 6.56 bpw | 839.68 GB | 7.05 GB | 864.82 GB | Ultra High Quality |
| GGUF Q5_K_M | 5.69 bpw | 728.32 GB | 7.05 GB | 751.23 GB | High Balanced Quality |
| Q4_K_M / AWQ | 4.85 bpw | 620.8 GB | 7.05 GB | 641.56 GB | Sweet Spot (Recommended) |
| GGUF IQ4_XS | 4.25 bpw | 544 GB | 7.05 GB | 563.23 GB | Compact 4-bit |
| GGUF Q3_K_M | 3.91 bpw | 500.48 GB | 7.05 GB | 518.84 GB | Moderate Perplexity Loss |
| GGUF Q2_K | 3.15 bpw | 403.2 GB | 7.05 GB | 419.61 GB | Extreme Compression |
🖥️ GPU & Apple Silicon Compatibility for Xiaomi MiMo-V2.6-Pro (1T MoE) (Q4_K_M)
Frequently Asked Questions
What are the exact architecture parameters of Xiaomi MiMo-V2.6-Pro (1T MoE)?
Xiaomi MiMo-V2.6-Pro (1T MoE) uses MiMoV2ForCausalLM with 70 layers (70-Layer Sparse MoE (384 Experts, Top-8 Active)), hidden_size=6144, 128 Q heads / 8 KV heads (head_dim=192), and vocab_size=152,576.
How much VRAM does Xiaomi MiMo-V2.6-Pro (1T MoE) require?
At 4-bit quantization (Q4_K_M, 16K context), Xiaomi MiMo-V2.6-Pro (1T MoE) requires 641.56 GB VRAM. Full FP16 requires 2097.31 GB.