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

Model details

MiMo-V2.6-Pro-UltraSpeed

MiMo-V2.6-Pro-UltraSpeed is Xiaomi's high-throughput variant of its flagship omnimodal model, rolled out alongside MiMo-V2.6-Pro and MiMo-V2.6-Flash as part of the same series release. Where Pro is positioned as Xiaomi's most capable model and Flash as the efficiency-tuned counterpart, UltraSpeed is specifically optimized for extreme generation speed, delivering up to 20x faster output at the same quality level. This makes it a practical fit for latency-sensitive workloads such as interactive agents, large-batch content generation, and real-time coding assistance where throughput matters more than absolute peak capability.

The UltraSpeed variant inherits the broader MiMo-V2.6 design philosophy of natively omnimodal input handling with text output, built within a publicly open-sourced series that Xiaomi describes as a step along an RSI path focused on scaling reinforcement learning compute on verifiable, complex tasks. On the coding benchmarks Xiaomi reports alongside the release, the underlying MiMo-V2.6-Pro posts competitive scores including 71.9 on DeepSWE v1.1 and 63.2 on the in-house MiMo Code Bench, giving UltraSpeed a strong quality baseline to draw from while its speed-focused optimizations take priority. In practical terms, it is best understood as a deployment-oriented sibling for teams that want frontier-class reasoning and coding ability without sacrificing responsiveness.

Xiaomimimo-v2.6-pro-ultraspeedmimo

Quick Info

Powered by
Provider
Xiaomi
Model key
mimo-v2.6-pro-ultraspeed
Release date
Sep 21, 2026
Last updated
Sep 21, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$4.35
Output token cost
$8.70

Limits

Output tokens
131,072 tokens
Context window
1,048,576 tokens

Transparent token rates

Compare MiMo-V2.6-Pro-UltraSpeed 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-V2.6-Pro-UltraSpeed

Xiaomi

Coverage

CellCog reports that Xiaomi released and open-sourced MiMo-V2.6 on September 21, 2026, shipping weights for Pro, Flash, and a 9B distillation on Hugging Face under MIT, alongside a public API on Xiaomi's platform and OpenRouter, and a public technical report. The article describes MiMo-V2.6-Pro as a 1.02-trillion-param The piece notes MiMo-V2.6-Pro scores 46 on Artificial Analysis' Intelligence Index, first among the 114 open-weight models tracked, level with Grok 4.7 and seven points behind Claude Fable 5.1 and GPT-6 Astra. Pricing for Pro is reported as unchanged from V2.5 at $0.435 per million input tokens and $0.87 per million ou

Xiaomi

Coverage

VentureBeat reports that Xiaomi released MiMo-V2.6-Pro and MiMo-V2.6-Flash on September 21, 2026, with MiMo-V2.6-Pro debuting as the top-performing open-weight model on Artificial Analysis' Intelligence Index at a score of 46. The piece places MiMo-V2.6-Pro ahead of proprietary models including xAI's Grok 4.6 (44) and The article details distribution and pricing: MiMo-V2.6-Pro is MIT-licensed and downloadable from Hugging Face, with API access priced at $0.435 per million uncached input tokens and $0.87 per million output tokens, while MiMo-V2.6-Flash is offered at $0.14/$0.28 per million input/output tokens with the same 1-million-

Xiaomi

Coverage

According to Xiaomi's official MiMo product page dated September 22, 2026, the company released and open-sourced the MiMo-V2.6 series as part of its "RSI path" exploration, scaling reinforcement-learning compute on verifiable, complex tasks. The series contains natively omnimodal models: MiMo-V2.6-Pro, the most capable The page reports Xiaomi's in-house benchmark scores across coding, agent, visual-coding, and cybersecurity suites (DeepSWE v1.1, ProgramBench, MiMo Code Bench, GDPVal 2.1 AA, Toolathlon-verified, Automation Bench v1.0.6, Agents' Last Exam, Terminal Bench 4.0, JobBench, Visual Coding, and CyberGym), placing MiMo-V2.6-Pr

Xiaomi

CoverageBenchmark

The LLMLearner model page specifically catalogs MiMo-V2.6-Pro-UltraSpeed as Xiaomi's high-speed serving tier for MiMo-V2.6-Pro, "offering up to 20x output speed at roughly 10x the standard price." Provider is listed as Xiaomi, model type as multimodal, released September 22, 2026. Specifications include 1.0T total para The page lists open weights under the MIT License with commercial use allowed, and recommends SGLang and vLLM serving recipes for self-hosting, noting that Pro requires multi-GPU serving while lighter distillations run on one card. No benchmark scores are reported on this aggregator page, and it lacks numerical confirm

Videos about MiMo-V2.6-Pro-UltraSpeed

More models around MiMo-V2.6-Pro-UltraSpeed