Typesafe · 105× · naposledy 27. 10. 2026
61
Momentum
Jev is the first public model from the new AI Lab TypeSafe AI and is referred to as a "System One Model" – a new model class for fast, structured decisions instead of text generation. Rather than generating text Token by Token, Jev answers typed questions in parallel and returns calibrated probabilities with confidence scores that can be directly processed by software. It was trained using a method developed by TypeSafe called "Reinforcement Learning for Calibrated Decisions" (RLCD); the model is currently in Early Access stage.
Vývoj momenta
02.08.31.10.
Vlastnosti
| Key-Benchmark (%) | 67,8% durchschnittliche Übereinstimmung auf TypeSafes internem 4-Workflow-Benchmark (vs. 74,1% beim besten Vergleichsmodell) |
| Kontextfenster (Token) | 64.000 Token pro Anfrage gesamt; max. 32.000 Token für Zustand plus längste Frage |
| Lizenz | Proprietär (nur API-Zugang, keine offenen Gewichte) |
| Multimodality | No text output; input is a 'state' (unstructured context) plus typed questions, output is typed values/probability distributions (no multimodal image/audio processing documented) |
| Platform | API/early access via TypeSafe AI (typesafe.ai), waitlist/Discord for developer access |
| Price per 1M Tokens | $42 per 1 billion input tokens (≈ $0.042 per 1M input tokens); output tokens free |
| Release Date | Launch day per official blog 'Introducing System One Models and Jev' (available in early access) |