The models · 46 open-weight · 24 labs
46 open-weight models, newest first.
The open-weight models the almanac covers. Entry is by the OpenRouter weekly token ranking plus explicit additions; a model leaving the ranking keeps its page, because the methods its documentation described do not expire.
Sort by any column. “Methods” counts what a model’s sources describe, so it measures how openly a lab writes about its model — a 200-method report and a 17-method model card can describe equally capable models. Sizes, context, attention family and licence are recorded by hand from the model card and config.
| Mistral-Large-4 | Mistral AI | 2026-10-29 | — | ~1T | ~52B | 1M | sparse (indexer) | 188 | 7 |
| LFM3-1.7B | Liquid AI | 2026-10-27 | — | 1.73B | dense | 256K | conv hybrid | 164 | 5 |
| MiniMax-M3.1-Flash | MiniMax | 2026-10-19 | #6 | ~428B | ~23B | 1M | sparse (indexer) | 11 | 3 |
| DeepSeek-V4.1-Flash-1017 | DeepSeek | 2026-10-17 | #1 | 552B | 16B | 1M | sparse (indexer) | 12 | 3 |
| NVIDIA-Nemotron-3.5-Super-VL-120B-A12B | NVIDIA | 2026-10-15 | — | 120B | 12B | 1M | Mamba hybrid | 277 | 7 |
| Step-5-Preview | StepFun | 2026-10-08 | #5 | 600B | 27B | 1M | sliding/global hybrid | 27 | 4 |
| Kolibri-1 | Aleph Alpha | 2026-10-03 | — | 78B | 3.46B | 256K | sliding/global hybrid | 410 | 4 |
| FLUX 3 Action | Black Forest Labs | 2026-09-23 | — | 7B | 7B | — | 24 | 3 | |
| MiMo-V2.6-Flash | Xiaomi | 2026-09-22 | #10 | 309B | 15B | 1M | sliding/global hybrid | 136 | 5 |
| MiMo-V2.6-Pro | Xiaomi | 2026-09-22 | — | 1.02T | 42B | 1M | sliding/global hybrid | 142 | 5 |
| DeepSeek-V4.1-Flash | DeepSeek | 2026-09-10 | #3 | 552B | 16B | 1M | sparse (indexer) | 105 | 4 |
| Hy4-preview | Tencent (Hunyuan) | 2026-08-28 | #4 | 770B | 49B | 1M | sparse (indexer) | 35 | 7 |
| GLM-5.3-Flash | Z.ai (Zhipu AI) | 2026-08-26 | #2 | 320B | 18B | 1M | linear hybrid | 33 | 5 |
| Qwen3.8-Flash-Next | Alibaba (Qwen) | 2026-08-26 | — | 125B | 6B | 256K | linear hybrid | 121 | 7 |
| DeepSeek-V4-Flash-Vision-Exp | DeepSeek | 2026-08-21 | — | 284B | 13B | 1M | sparse (indexer) | 25 | 3 |
| GLM-5.3 | Z.ai (Zhipu AI) | 2026-08-16 | #7 | ~753B | ~40B | 1M | sparse (indexer) | 46 | 6 |
| DeepSeek-V4-Pro-0813 | DeepSeek | 2026-08-13 | — | 1.6T | 49B | 1M | sparse (indexer) | 13 | 3 |
| NVIDIA-Nemotron-3.5-Lightning-30B-A3B | NVIDIA | 2026-08-11 | — | 30B | 3B | 1M | Mamba hybrid | 284 | 8 |
| DeepSeek-V4-Flash-0731 | DeepSeek | 2026-07-31 | #8 | 284B | 13B | 1M | sparse (indexer) | 17 | 4 |
| Xiaomi-Robotics-1 | Xiaomi | 2026-07-28 | — | — | — | — | 46 | 3 | |
| Laguna-S-2.1 | Poolside | 2026-07-21 | — | 118B | 8B | 1M | sliding/global hybrid | 163 | 8 |
| Kimi K3 | Moonshot AI | 2026-07-16 | — | 2.8T | 104B | 1M | linear hybrid | 185 | 10 |
| Inkling | Thinking Machines Lab | 2026-07-15 | — | 975B | 41B | 1M | sliding/global hybrid | 72 | 11 |
| LingBot-VLA 2.0 | Robbyant (Ant Group) | 2026-07-07 | — | 6.4B | 6.4B | — | 91 | 5 | |
| Hy3 | Tencent (Hunyuan) | 2026-07-06 | #9 | 295B | 21B | 256K | full (GQA/MHA) | 37 | 7 |
| GLM-5.2 | Z.ai (Zhipu AI) | 2026-06-16 | — | ~753B | ~40B | 1M | sparse (indexer) | 68 | 5 |
| NVIDIA-Nemotron-3-Ultra-550B-A55B | NVIDIA | 2026-06-04 | #11 | 550B | 55B | 1M | Mamba hybrid | 205 | 7 |
| MiniMax-M3 | MiniMax | 2026-06-01 | — | ~428B | ~23B | 1M | sparse (indexer) | 81 | 7 |
| Step-3.7-Flash | StepFun | 2026-05-29 | — | ~196B | ~11B | 256K | sliding/global hybrid | 43 | 6 |
| MolmoAct 2 | Allen Institute for AI | 2026-05-04 | — | 5.4B | 5.4B | — | 165 | 3 | |
| DeepSeek-V4-Flash | DeepSeek | 2026-04-24 | #12 | 284B | 13B | 1M | sparse (indexer) | 82 | 4 |
| DeepSeek-V4-Pro | DeepSeek | 2026-04-24 | — | 1.6T | 49B | 1M | sparse (indexer) | 97 | 5 |
| MiMo-V2.5 | Xiaomi | 2026-04-22 | — | 310B | 15B | 1M | sliding/global hybrid | 76 | 8 |
| MiMo-V2.5-Pro | Xiaomi | 2026-04-22 | — | 1.02T | 42B | 1M | sliding/global hybrid | 42 | 6 |
| GR00T N1.7 | NVIDIA | 2026-04-17 | — | 3.1B | 3.1B | — | 82 | 4 | |
| VLA Foundry (Foundry-VLA-1.7B) | Toyota Research Institute | 2026-04-17 | — | 1.7B | 1.7B | — | 68 | 3 | |
| Qwen3.6-35B-A3B | Alibaba (Qwen) | 2026-04-15 | — | 35B | 3B | 256K | linear hybrid | 46 | 6 |
| Gemma 4 31B | Google DeepMind | 2026-04-02 | — | 30.7B | dense | 256K | sliding/global hybrid | 79 | 10 |
| Alpamayo 1.5 / 2 Super | NVIDIA | 2026-03-03 | — | 10B | 10B | — | 61 | 2 | |
| Qwen3.5-397B-A17B | Alibaba (Qwen) | 2026-02-16 | — | 397B | 17B | 256K | linear hybrid | 100 | 8 |
| DeepSeek-V3.2 | DeepSeek | 2025-12-01 | — | 671B | 37B | 160K | sparse (indexer) | 68 | 4 |
| X-VLA | X-VLA authors | 2025-11-04 | — | 0.9B | 0.9B | — | 50 | 2 | |
| π0.5 | Physical Intelligence | 2025-09-09 | — | 3.6B | 3.6B | — | 107 | 5 | |
| gpt-oss-120b | OpenAI | 2025-08-05 | — | 117B | 5.1B | 128K | sliding/global hybrid | 84 | 6 |
| SmolVLA | Hugging Face | 2025-06-03 | — | 0.45B | 0.45B | — | 62 | 3 | |
| OpenVLA / OpenVLA-OFT | OpenVLA project (Stanford et al.) | 2024-06-10 | — | 7.5B | 7.5B | — | 73 | 3 |
Rank is the model's position among open-weight models in OpenRouter's token ranking for the week of 2026-10-26, the same week for every model; — means outside that week's top 12. Params and Active are total and per-token parameters; hover an attention family for its detail.