AI Models & Research
New language and multimodal models, open weights, training and inference techniques, and papers worth reading.
Rising now
Open the radar| # | Topic | 14-day trend | Channels in 7 days | Score |
|---|---|---|---|---|
| 1 | Qwen | 8 ch12 voices | 93 | |
| 2 | DeepSeek | 4 ch6 voices | 68 | |
| 3 | Gemma | 4 ch4 voices | 68 | |
| 4 | Llama | 4 ch5 voices | 67 | |
| 5 | lm_understand-logs | 2 ch2 voices | 45 | |
| 6 | Awesome-Omnimodal-Embeddings | 2 ch2 voices | 44 | |
| 7 | lab2sim | 2 ch2 voices | 41 | |
| 8 | vLLM | 2 ch2 voices | 34 | |
| 9 | Kimi (Moonshot AI) | 2 ch2 voices | 26 | |
| 10 | unslop | 1 ch1 voices | 25 |
Gemma
Gemma 4 Open Models: What to Check Before Running Them Locally
Open-weight multimodal Gemma models now span 2.3B to 31B parameters. Here is where they fit in creative and dev work, with the reported evidence and limits.
Qwen
Qwen 3.8 Is Running Hot: What Creators Should Check First
Why Qwen3.8, Qwen-Image-2.1 and coding agents surged at once, and what to verify before wiring them into design, video and 3D work.