Choose a model to help preserve. Explore the archive shortlist and the Artificial Analysis ranking, with a magnet when one is listed.
Cement yourself in computing history by seeding this shortlist of landmark sovereign AI models, so future users can explore what these models did, how they worked, and what people built with them.
| Model | 🤗 downloads | Est. size | Seeders | Magnet |
|---|---|---|---|---|
 DeepSeek-V4.1-Flash-1017deepseek-ai/DeepSeek-V4.1-Flash-1017 · mit | 434k | 510 GB | 🌱 1 | Magnet |
 Qwen3.5-35B-A3BQwen/Qwen3.5-35B-A3B · apache-2.0 | 1.5M | 72 GB | 🌱 18 | Magnet |
| Qwen3.8-27BQwen/Qwen3.8-27B · apache-2.0 | 6.8M | 56 GB | 🌱 118 | Magnet |
| 6.4M | 328 GB | 🌱 45 | Magnet | |
| 804k | 1.6 TB | 🌱 4 | Magnet | |
| 1.7M | 510 GB | 🌱 45 | Magnet |
An editorial selection for archival purposes, not a benchmark ranking.
Artificial Analysis runs independent evaluations across reasoning, coding, knowledge and agentic tasks, with a published methodology that makes the basis of comparison transparent.
| # | Model | AA Index | Est. size | Seeders | Magnet |
|---|---|---|---|---|---|
| 1 | 46.5 | 1.6 TB | 🌱 9 | Magnet | |
| 2 | 32.6 | 510 GB | 🌱 1 | Magnet | |
| 3 | 30.6 | 394 GB | 🌱 0 |
| Magnet |
| 4 |  MiMo-V2.6-ProXiaomiMiMo/MiMo-V2.6-Pro-RL · mit | 29.9 | 573 GB | 🌱 2 | Magnet |
| 5 |  Ling 3.1 Flash 1019inclusionAI/Ling-3.1-flash-1019 · mit | 29.6 | 574 GB | 🌱 12 | Magnet |
| 6 |  Mistral Large 4 (High)mistralai/Mistral-Large-4-1T-A52B · other | 26.1 | 1.1 TB | 🌱 0 | Magnet |
| 7 | Qwen3.8 27B (Xhigh)Qwen/Qwen3.8-27B · apache-2.0 | 20.1 | 56 GB | 🌱 118 | Magnet |
| 8 |  K2 Horizon 375B A23BIFM/K2-Horizon-375B-A23B · apache-2.0 | 17.8 | 758 GB | 🌱 9 | Magnet |
| 9 |  Beam 1.0reflection/Beam-1.0-501B-A23B · apache-2.0 | 17.7 | 1.0 TB | 🌱 0 | Magnet |
| 10 |  Nemotron 3.5 Super VLnvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 · openmdw-1.1 | 13.6 | 249 GB | 🌱 11 | Magnet |
