Decision modelOpen source
Kev 0.8B
Jared Palmer
- Released
- September 20, 2026
- Data date
- September 24, 2026
- Access
Kev 0.8B is the smallest current Kev adapter variant. Jared Palmer publishes it as a LoRA adapter with a pointer head on a 0.8-billion-parameter base model. The local checkpoint reads text or structured states and typed questions, then returns Choice, Noul, and Score with calibrated probabilities in one pass.
Kev 4B and Kev 9B use the same pointer-head idea with larger base models. All three combine an adapter with a base model and have a TypeSafe-compatible interface.
Specifications and access
| Specification | Value and source |
|---|---|
| Model class | Typed-decision LoRA adapter with pointer readoutSource |
| Access | LocalSource |
| Model ID | jaredpalmer/kev-0.8bSource |
| Input | Text or structured states and typed questionsSource |
| Output | Structured probabilities, no text generationSource |
| Result | Choice, Noul and Score with calibrated probabilitiesSource |
| Parameters | 0.8 billion base modelSource |
| License | Apache-2.0 for the adapter and headSource |
| Scope | Local Kev family models with a TypeSafe-compatible interface.Source |
| Notes | The weights are adapters on top of the respective Qwen base models.Source |
| Base model | Qwen/Qwen3.5-0.8B-BaseSource |
| Adapter | LoRA, rank 16Source |
Popularity
42.9 / 100
Sources and data date
| Type | Evidence and data date |
|---|---|
| Additional source | Kev repository (retrieved September 24, 2026) · Kev repository · Editorial description reviewed October 4, 2026 |
| Additional source | Kev release plan (retrieved September 24, 2026) · Kev release plan · Editorial description reviewed October 4, 2026 |
| Additional source | Kev 0.8B checkpoint (retrieved September 24, 2026) · Kev 0.8B checkpoint · Editorial description reviewed October 4, 2026 |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Creative Commons Attribution 4.0 |