MiMo-V2.5 310B — VRAM, speed & local setup
Written by Jakub Rusinowski · Last updated
310B total, 15B active per token, MIT, 1M context. Every one of the 310B must be resident — 188 GB at Q4_K_M, so a 192 GB Mac Studio or a small multi-GPU box — while per-token bandwidth is that of a 15B model. The standard variant rather than the flagship, and the cheaper of the two to serve.
MiMo-V2.5 310B needs about 188 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
VRAM and speed by quantization
Quoted against NVIDIA RTX 4090 (24 GB). Weights plus overhead only: this model's architecture is unpublished, so no KV cache is added.
| Quant | Memory | VRAM | Speed (est.) | Fit |
|---|---|---|---|---|
| Q2_K 2.63 bpw | 102.7 GB | — | Too big | |
| Q3_K_M 3.41 bpw | 132.9 GB | — | Too big | |
| Q4_K_M 4.83 bpw | 188 GB | — | Too big | |
| Q5_K_M 5.67 bpw | 220.5 GB | — | Too big | |
| Q6_K 6.56 bpw | 255 GB | — | Too big | |
| Q8_0 8.50 bpw | 330.2 GB | — | Too big | |
| F16 16.00 bpw | 620.8 GB | — | Too big |
Black marker = usable memory on the NVIDIA RTX 4090 (24 GB). Estimates from the memory-bandwidth roofline on the methodology page. MiMo-V2.5 310B VRAM calculator →
Get MiMo-V2.5 310B running
The cheapest catalogued GPU that runs MiMo-V2.5 310B is the Apple M3 Ultra (512 GB).
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
How to run MiMo-V2.5 310B
Install Ollama, then run:
ollama run mimo-v2-5Specifications
Verified — Checked against the primary source — the model card or the vendor spec page — and corroborated by a second independent source.
- Parameters
- 310 Billion (15B active)
- Context window
- 1,000,000
- Architecture
- Mixture-of-Experts
- Provider
- Xiaomi
- Licence
- MIT
- Practical deployment
- Workstation (large unified memory or pro GPU)
- Specified at
- Q4_K_M
- System RAM
- 256 GB
- Record updated
- 2026-09-06
Commercial use permitted. No usage restrictions beyond attribution.
Quality and use cases
Scores as published by the model’s authors or an independent evaluator — quality, not throughput, and not measured by us.
Can I run MiMo-V2.5 310B on my GPU?
- MiMo-V2.5 310B on AMD Ryzen AI Max+ 395
- MiMo-V2.5 310B on Apple M1 Ultra
- MiMo-V2.5 310B on Apple M2 Max
- MiMo-V2.5 310B on Apple M2 Ultra
- MiMo-V2.5 310B on Apple M4 Max
- MiMo-V2.5 310B on Apple M5 Max
- MiMo-V2.5 310B on NVIDIA A100 80GB (PCIe)
- MiMo-V2.5 310B on NVIDIA DGX Spark
- MiMo-V2.5 310B on NVIDIA H100 80GB (PCIe)
- MiMo-V2.5 310B on NVIDIA RTX PRO 6000 Blackwell
Other MiMo-V2.5 sizes
MiMo-V2.5 310B — frequently asked questions
How much VRAM does MiMo-V2.5 310B need?
About 188 GB at Q4_K_M — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does MiMo-V2.5 310B run on an RTX 4090 (24 GB)?
No. MiMo-V2.5 310B needs about 188 GB at Q4_K_M, more than a single RTX 4090's 24 GB. It needs a larger card, several GPUs, or Apple Silicon with enough unified memory — or it runs with part of the weights offloaded to system RAM, which is much slower.
How do I run MiMo-V2.5 310B locally?
Install Ollama and run `ollama run mimo-v2-5`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.
What other sizes does MiMo-V2.5 come in?
MiMo-V2.5 310B (188 GB), MiMo-V2.5-Pro 1T (605 GB). Every size shares the family's training and licence; the larger ones score higher and need proportionally more memory.