Kev
An open decision model family: it answers typed questions (yes/no, choice, rating) with a probability for each answer instead of generating text. Kev 1.0 has four sizes: 0.8B, 4B and 9B adapters on Qwen3.5 base models, and a fully fine-tuned 27B on Qwen3.8. The server matches TypeSafe's System One API, so the TypeSafe Python SDK works against a self-hosted Kev. Documents up to 65,536 tokens on the 27B; the smaller sizes are validated to 8,192.
No host has a verified price for Kev yet. The weights are open, so you can run it yourself.
Every decision model with its hosts and prices is on the decision models page. For text-generating models, see LLM API pricing by model.
Compare every inference provider on free tiers, hosting and EU availability in the inference APIs table.
Is your product missing?