Model Families: Variants, Hardware Requirements and Coverage
Model families covered by Local Model Watch with more than one article or with converted builds tracked after publication. Each page gathers the family’s articles, hardware requirements computed by this site, and quantized / converted variants. For a single table of all models by GPU memory, see the VRAM quick reference.
To browse by the company or lab that released the models, see the publisher index; to browse by the file format (GGUF, MLX, FP8 …), see the model format index.
Last updated 2026-11-02 (JST). 17 families.
→ Scroll horizontally to see all columns
| Family | Publisher | Parameters | Smallest VRAM tier | Articles | Variants tracked | Latest |
|---|---|---|---|---|---|---|
| Qwen3.8 | Alibaba (Qwen) | 27.8B | 8GB | 8 | 16 | 2026-10-02 |
| MiniMax-H3 | MiniMax | — | 32GB | 6 | 2 | 2026-10-06 |
| MiMo-V2.6 | Xiaomi (MiMo) | — | — | 4 | 4 | 2026-09-28 |
| YuE2 | M-A-P | 3.6B | 12GB | 3 | 0 | 2026-09-15 |
| K2-Horizon | IFM (MBZUAI) | — | 32GB | 3 | 1 | 2026-09-23 |
| Qwen-Image-2.1 | Alibaba (Qwen) | — | — | 3 | 0 | 2026-09-30 |
| Nex-N2.5 | Nex-AGI | — | — | 2 | 6 | 2026-09-09 |
| DeepSeek-V4 | DeepSeek | 1650.5B | — | 2 | 3 | 2026-09-26 |
| MiniCPM5 | OpenBMB | 2.5B | 4GB | 1 | 5 | 2026-09-08 |
| DeepSeek-V4.1 | DeepSeek | 763.2B | — | 1 | 1 | 2026-09-10 |
| WeVisDoc | Tencent | 4.4B | 4GB | 1 | 2 | 2026-09-18 |
| plamo-3-fin | Preferred Networks (PFN) | 890M | 4GB | 1 | 1 | 2026-09-24 |
| LFM2.5-VL | Liquid AI | — | — | 1 | 1 | 2026-09-25 |
| Xing4.0 | China Telecom (Xing, TeleChat) | 31.2B | 24GB | 1 | 3 | 2026-09-28 |
| LTX-2.5 | Lightricks (LTX) | — | — | 1 | 1 | 2026-10-01 |
| gemma-4-E4B | 8.0B | 8GB | 1 | 1 | 2026-10-05 | |
| Naive-N0.5 | NaiveAI | — | — | 1 | 2 | 2026-10-05 |