Models
Model pages identify an open-weight checkpoint: developer, parameter count when the card states it, modalities, license string (including the opaque Hugging Face label “other”), and likes as of the verification date. We do not invent VRAM figures or leaderboard scores. Hardware rows that say “not yet verified” are empty on purpose.
Like tools, each model is attached to an alternatives hub — for example open embeddings versus a closed embedding API, or a decision model versus a classification API. That mapping is editorial. For head-to-heads that stay inside sourced facts, see comparisons.
| Model | Type | Developer | Parameters | License |
|---|---|---|---|---|
| Qwen-Image-2.1-Turbo | image-gen | Qwen | Not yet verified | other |
| Kolibri-1 | llm | Aleph Alpha | 78.1B | Apache-2.0 |
| Cloudflare Clef | decision | Cloudflare | 27B | Apache-2.0 |
| EmbeddingGemma 2 | embedding | Google DeepMind | ~0.74B | Apache-2.0 |
| BGE-M3 | embedding | BAAI | Not yet verified | MIT |