# Jev, Explained > What TypeSafe AI's Jev decision model is: typed answers instead of text, the three question types (Choice, Score, Noul), pricing at $0.042 per 1M input tokens with free output, and how it compares to LLMs, OpenAI's Decisions API, and AWS Strands Decider 2B. Includes a cost calculator. Independent explainer, not affiliated with TypeSafe. Full page: https://lab.awaara.cn/jev/ ## What Jev is - Jev is an AI model from TypeSafe AI (San Francisco, founded 2024 by Diogo Almeida, Erik Gafni, Sasha Sheng). Released September 15, 2026. - It does not generate text. You send it a "state" (text or JSON) plus typed questions, and it returns structured answers: a picked option, a score, or a yes/no probability. - Three question types: Choice (pick one from up to 255 options), Score (rate against ordered levels), Noul (yes/no, one probability 0-1). - Prefill-only: one parallel pass through the transformer, 70-500 ms per decision (TypeSafe's figure). Asking ten questions about one state costs barely more than asking one. - Trained with RLCD (reinforcement learning for calibrated decisions): probabilities are meant to be honest, e.g. 0.8 means right ~80% of the time. - Pricing: $0.042 per 1M input tokens ($42 per 1B), output free. No open weights, no architecture paper. Early-access waitlist on typesafe.ai; Python/JS/TS SDKs; playground and console API keys. - Weak spots (TypeSafe's "jaggedness" list): reads literally, bad at math/counting, struggles with multi-step reasoning, text only (English best), no chain-of-thought. - Alternatives: OpenAI Decisions API (Sep 29, 2026; specialized GPT-6 Luna; $0.10/1M input, output free; accepts images; self-reported confidence), AWS Strands Decider 2B (Oct 1, 2026; open source Apache 2.0; runs on your CPU/GPU). - October 9, 2026: TypeSafe raised $870M at a $7.5B valuation (TechCrunch). Adoption and speed figures quoted are the company's own; no independent benchmarks published yet. ## Links - Explainer: https://lab.awaara.cn/jev/ - Cost calculator: https://lab.awaara.cn/jev/#calc - Related: OpenAI Decisions API explained https://lab.awaara.cn/decisions-api/ ยท Haiku 5.5 pricing https://lab.awaara.cn/claude-haiku-5-5/