MiMo-V2.5-Pro-UltraSpeed
Unavailable in this round
Language modelPreview
MiMo-V2.5-Pro-UltraSpeed
Xiaomi
- Released
- June 2026
- Data date
- September 8, 2026
- Price
- Input · Output / 1M tokens
MiMo-V2.5-Pro-UltraSpeed is Xiaomi’s accelerated serving variant of MiMo-V2.5-Pro. The parameter figures belong to the underlying Pro model. Xiaomi describes TileRT and DFlash as the technical basis and reports more than 1,000 output tokens per second on a particular eight-GPU setup.
Xiaomi lists the model as a preview, and the speed figure is a vendor claim. The initial internal trial concluded, but on September 8, 2026 Xiaomi extended the limited-time chat and API experience. Approved users could continue using the closed beta; the specific offline time was still open. UltraSpeed belongs to the serving layer of MiMo-V2.5-Pro, while quality measurements remain attributable to the underlying Pro model.
Specifications and access
| Specification | Value and source |
|---|---|
| Weights | Open weightsSource |
| Access | API, LocalSource |
| Parameters | 1TSource |
| Active parameters | 42BSource |
| Context window | 1,000,000 tokensSource |
| Notes | Accelerated serving variant of MiMo-V2.5-Pro. Parameter counts refer to the underlying Pro model. Xiaomi's speed figure is a vendor claim, not an independent measurement.Source |
| Architecture | Accelerated serving of MiMo-V2.5-Pro using TileRT and DFlash. DFlash drafts multiple tokens for the main model to verify. (retrieved 2026-09-22)Source |
| Quantization | MXFP4 for MoE experts. Other modules retain their original precision. (retrieved 2026-09-22)Source |
| Throughput in tokens per second | Xiaomi reports more than 1,000 output tokens/s on an eight-GPU node. Not an independent measurement or guaranteed API speed. (retrieved 2026-09-22)Source |
| Availability | The official notice says the limited trial has ended and announces a forthcoming commercial version. (retrieved 2026-09-22)Source |
Our tests
SVG short test
One SVG prompt and fixed output settings. Captures are shown only with 0 reported reasoning tokens or documented lack of reasoning controls. Identical test conditions do not make the models' internal compute identical.
Popularity
28.9 / 100
Sources and data date
| Type | Evidence and data date |
|---|---|
| Price | Base API prices without caching or batch discounts. Higher context tiers and other rates are listed under Costs. |
| Research status | Research date September 8, 2026. 3 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not found in the inspected sources | parameters-billions Original wording No exact variant-specific value was located; the available page describes a related/base model. |
| Not found in the inspected sources | release-date Original wording No exact UltraSpeed release date was located; the checked official page documents the related MiMo-V2.5-Pro weights. |
| Not found in the inspected sources | active-parameters-billions Original wording No exact variant-specific value was located; the available page describes a related/base model. |
| Not found in the inspected sources | knowledge-cutoff Original wording No exact variant-specific value was located; the available page describes a related/base model. |
| Not found in the inspected sources | profile-benchmarks Original wording Checked the exact primary model card/API or release page; no usable labeled benchmark score was published/found for this model. |
| Unresolved | variant-specific-benchmarks Original wording The checked official material describes MiMo-V2.5-Pro weights; no separate UltraSpeed evaluation was found. |
| Unresolved | source-discrepancy Original wording The official MiMo page and HF card document MiMo-V2.5-Pro weights; no separate UltraSpeed benchmark/model card was found, so Pro results are not copied to this variant. |
| Unresolved | AA benchmark rows Original wording No exact Artificial Analysis model page located in the checked public sitemap. |
| Additional source | Model metadata (retrieved September 22, 2026) · Xiaomi UltraSpeed announcement · Editorial description reviewed October 2, 2026 |
| Additional source | Xiaomi UltraSpeed beta extension notice · Editorial description reviewed October 2, 2026 |
| Additional source | MiMo-V2.5-Pro model card · Editorial description reviewed October 2, 2026 |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Creative Commons Attribution 4.0 |