Vendor · Language models
Xiaomi has 8 models in the directory. The strongest is MiMo V2.6 Pro: Rank 20 to 31 of 95 in the overall index.
In the directory: 8 language models. 7 of them with open weights that can run locally.
Every model there ever was. Models with a profile lead to every benchmark, price and source.
| MiMo V2.6 Distill Qwen 9B | Sep 22, 2026 | Open weights | n/a | n/a | n/a | n/a | 46 |
| MiMo V2.6 Flash | Sep 22, 2026 | Active | 41Rank 27 to 38 of 95 | $0.14 / $0.28 | 56 | n/a | 34 |
| MiMo V2.6 Pro | Sep 22, 2026 | Active | 51Rank 20 to 31 of 95 | $0.44 / $0.87 | 58 | n/a | 16 |
| MiMo V2.6 Pro UltraSpeed | Sep 22, 2026 | Active | n/a | $4.35 / $8.70 | n/a | n/a | below 1 |
| MiMo-V2.5-Pro-UltraSpeed | June 2026 | Preview | n/a | $1.31 / $2.61 | n/a | 1M | 29 |
| MiMo-V2.5-Pro | Apr 27, 2026 | Open source | 3Rank 55 to 70 of 95 | $0.44 / $0.87 | 51 | 1M | 40 |
| MiMo-V2.5 | Apr 23, 2026 | Open source | 2Rank 62 to 70 of 95 | $0.14 / $0.28 | 72 | 1M | 45 |
| MiMo-V2-Flash | Dec 16, 2025 | Active | n/a | n/a | n/a | 262K | 34 |
Language model · Sep 22, 2026
MiMo V2.6 Distill Qwen 9B is Xiaomi’s local 9B SFT checkpoint fine-tuned from Qwen 3.5 9B on MiMo-generated data. It accepts text and images and is much smaller than the full V2.6 Pro and Flash RL models. The checkpoint is aimed less at a hosted API and more at research and custom inference.
Language model · Sep 22, 2026
MiMo V2.6 Flash is Xiaomi’s smaller V2.6 class for frequent, cost-sensitive agent calls. The open Flash RL release uses a sparse MoE base of about 309B parameters with 15B active. The line remains multimodal and built for long contexts and reasoning.
Language model · Sep 22, 2026
MiMo V2.6 Pro is Xiaomi’s large V2.6 configuration for multimodal reasoning, coding, and agent work. Xiaomi deliberately gives the API service and the local MiMo-V2.6-Pro-RL checkpoint different names. The model ID therefore determines whether the service or the local RL checkpoint is meant.
Language model · Sep 22, 2026
MiMo V2.6 Pro UltraSpeed is Xiaomi’s accelerated API and desktop mode for MiMo V2.6 Pro. Xiaomi refers to the regular Pro checkpoint for the released weights. The name describes lower-latency serving within the V2.6 family.
Language model · Apr 27, 2026
MiMo-V2.5-Pro moves the open MiMo line to a sparse MoE base of about 1T parameters, with 42B active. Like MiMo-V2.5, Pro handles multiple modalities in a 1M-token context. Xiaomi therefore separates it technically and as a product from the standard model.
Language model · Apr 23, 2026
MiMo-V2.5 brings text, images, video, and audio into Xiaomi’s openly available 1M-token checkpoint. Its model card describes the hybrid attention of MiMo-V2-Flash as the architectural basis. The Pro variant remains a separate tier for harder agent and coding work.
Language model · Dec 16, 2025
MiMo-V2-Flash is Xiaomi’s text-only sparse MoE base with about 309B parameters, of which 15B are active per token. The openly available model supports up to 262,144 context tokens and up to 65,536 output tokens. Image and video input are explicitly outside this checkpoint.
Derived from the overall index, the API price and the measured speed of the current models.
The strongest choice
MiMo V2.6 Pro
The vendor's highest overall index
Value for money
MiMo V2.6 Flash
Costs 32% of MiMo V2.6 Pro, overall index 41 instead of 51
The fastest
MiMo-V2.5
72 tokens/s, a typical model manages 98
The 20 best language models, with the models from Xiaomi highlighted. “All” and the filter show every model in the index.
20 of 95 models
Provider
Weights
Price (blended per 1M tokens)
Released
Anthropic
OpenAI
Meta
Moonshot AI
DeepSeek
xAI
Z.ai
Xiaomi
Alibaba
StepFun
Mistral AI
MiniMax
Thinking Machines Lab
Inception
| Vendor | Models | Model types | Best language model | Top 10 | Open weights | New in 2026 |
|---|---|---|---|---|---|---|
| Anthropic | 33 | 33 language models | Claude Opus 5.5 (Rank 1 to 3) | 4 | 0 | 13 |
| OpenAI | 63 | 46 language models, 4 image models, 2 video models, 9 audio models, 2 specialist models | GPT-6 Astra (Rank 4 to 8) | 2 | 4 | 27 |
| 66 | 43 language models, 4 image models, 6 video models, 11 audio models, 2 specialist models | Gemini 4 Carbon (Rank 1 to 5) | 2 | 19 | 26 | |
| Meta | 32 | 29 language models, 1 image model, 1 video model, 1 specialist model | Muse Wonder Full (Rank 6 to 9) | 2 | 26 | 7 |
| DeepSeek | 23 | 23 language models | DeepSeek V4.1 Flash 1017 (Rank 13 to 20) | 0 | 23 | 9 |
| Moonshot AI | 12 | 12 language models | Kimi K3.1 Pro (Rank 10 to 13) | 0 | 10 | 7 |
| xAI | 16 | 12 language models, 2 image models, 2 video models | Grok 4.7 (Rank 18 to 24) | 0 | 1 | 12 |
| Xiaomi | 8 | 8 language models | MiMo V2.6 Pro (Rank 20 to 31) | 0 | 7 | 7 |
By release date. Grey: retired models.
7 releases since January 2026, by release date.
Prices per 1M tokens for the current language models, according to Xiaomi or the provider linked in each model profile.
| Current model | Input | Output | Context |
|---|---|---|---|
| MiMo V2.6 Pro | $0.44 | $0.87 | n/a |
| MiMo V2.6 Flash | $0.14 | $0.28 | n/a |
| MiMo-V2.5-Pro | $0.44 | $0.87 | 1M |
| MiMo-V2.5 | $0.14 | $0.28 | 1M |
| MiMo-V2.5-Pro-UltraSpeedPreview | $1.31 | $2.61 | 1M |
| MiMo V2.6 Pro UltraSpeed | $4.35 | $8.70 | n/a |
The directory lists 8 MiMo models from Xiaomi: 8 language models. It counts every model Xiaomi ever released, including retired ones. 7 are currently available, the newest being MiMo V2.6 Distill Qwen 9B, MiMo V2.6 Flash and MiMo V2.6 Pro.
MiMo V2.6 Pro. It is at rank 20 to 31 of 95 in the overall index, which combines independent benchmarks from several sources.
MiMo V2.6 Distill Qwen 9B, released on Sep 22, 2026. The timeline below shows every release by date.
Through the API per 1M tokens (input / output): MiMo V2.6 Flash $0.14 / $0.28 and MiMo V2.6 Pro $0.44 / $0.87. The prices of all 6 current language models are in the price table.
In the overall index, no. The best OpenAI model, GPT-6 Astra (rank 4 to 8 of 95), is ahead of MiMo V2.6 Pro (rank 20 to 31 of 95). Single tasks can differ.
Yes, 7 of the 8 models have open weights. Each model profile lists how much memory it needs.
MiMo-V2.5 with 72 tokens per second according to Artificial Analysis. A typical model manages 98.
Up to 1 M tokens (≈ 1,500 A4 pages), which 3 of the current models reach, including MiMo-V2.5-Pro-UltraSpeed. Every model's value is in the model list.
Data as of Nov 1, 2026