MiMo-V2.6 — local AI model by Xiaomi
Written by Jakub Rusinowski · Last updated
Xiaomi's September 2026 flagship and the successor line to MiMo-V2.5 and MiMo-V2-Pro. One released checkpoint, MiMo-V2.6-Pro: a 1.02-trillion-parameter sparse MoE with 42B active per token that takes text, image, video and audio in, with a 1M-token context, under MIT. Every expert must be resident, so this is a datacenter model: Xiaomi's own serving recipes are SGLang and vLLM on multi-GPU servers.
Variants
The smallest MiMo-V2.6 variant needs about 619 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| MiMo-V2.6-Pro → 1.02T (42B active) | ~619.2 GB | 1,048,576 | cloud-hosted tag |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run MiMo-V2.6 locally
Install Ollama, then pull the tag.
No Ollama library entry. Xiaomi's documented serving routes are SGLang and vLLM.
Pick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: MiMo-V2.6-ProMiMo-V2.6 — frequently asked questions
How much VRAM does MiMo-V2.6 need?
MiMo-V2.6 needs about 619 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: MiMo-V2.6-Pro (619 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run MiMo-V2.6 on an RTX 4090 (24 GB)?
MiMo-V2.6's smallest variant needs about 619 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
What quantization should I use for MiMo-V2.6?
Q4_K_M is the best balance of quality and VRAM for MiMo-V2.6 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run MiMo-V2.6 with Ollama?
MiMo-V2.6 has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.