Giải thích mô hình AI ra quyết định: TypeSafe Jev so sánh với Fastino GLiDE, GLiNER2.5-Decide và các đối thủ nguồn mở

Decision AI models are a new class of model that returns a decision, not a paragraph. You send text with typed questions. The model returns choices, scores or yes/no probabilities your code can branch on directly.

The category went mainstream when TypeSafe AI launched Jev：https://typesafe.ai/blog/introducing-system-one-models-and-jev after 2 years in stealth. TypeSafe calls it a ‘System One model,’ after Daniel Kahneman’s fast, intuitive System 1 thinking.

Within 3 weeks, Fastino Labs shipped 2 rival models, and open-source developers published several Jev-style reproductions. This article covers how the category works, where it fits, and how the options compare.

Jev accepts a ‘state’ (a string, array or set of name-value pairs) and one or more questions. According to the TypeSafe docs：https://docs.typesafe.ai/introduction, it supports 3 primitives:

Every question is evaluated in parallel and in isolation against the same state. Adding questions barely changes response time. Because Jev never generates strings, TypeSafe says it cannot return a type error.

Under the hood, TypeSafe describes a new architecture, a parallel sampler and a training method called Reinforcement Learning for Calibrated Decisions (RLCD) . RLHF optimizes for human preference. RLCD optimizes for calibrated probabilities, where higher confidence should mean higher accuracy.

Pricing is the key thing to watch. Jev costs $0.042 per million input tokens , and output is free. OpenRouter：https://openrouter.ai/typesafe/jev-1.13 lists a 32K context window. TypeSafe reports end-to-end responses in 70 to 500 milliseconds.

TypeSafe built workflow evals：https://evals.typesafe.ai/ across 4 tasks: security incidents, agent trace observability, invoice processing and customer service. Reference labels come from averaging GPT-6 Astra and Claude Fable 5.1 at high thinking.

Jev matched Sonnet 5 on accuracy at a fraction of the cost and latency. It still trails the top frontier configuration by 6.3 points. Per task, Jev scored 76.0% on customer service but only 61.8% on invoice processing.

The rule of thumb is simple. If your code needs a bounded answer it will branch on, a decision model is a candidate. If a human needs to read the output, use an LLM.

Vercel reports：https://vercel.com/blog/ai-gateway-jev-model-launch that Jev became the fastest-adopted model in AI Gateway history. Within 24 hours, nearly 13% of paid teams were using it. That was 2x the GPT-5.6 family’s share and more than 6x Fable 5.1’s.

Simon Willison fetched 100 candidates with BM25, then had Jev score each for relevance. This pattern replaces an expensive LLM reranker with cheap, parallel Score questions.

Arize：https://arize.com/docs/ax/evaluate/jev-as-a-judge and Langfuse：https://langfuse.com/docs/evaluation/evaluation-methods/jev-as-a-judge both shipped Jev-as-a-judge evaluators. Langfuse labels the feature “decision-model evaluators” so other models can be added later. Its Jev support is still marked experimental.

A new arXiv paper：https://arxiv.org/pdf/2609.23136 used Jev to interpret service contracts at the network edge. At matched correctness, Jev cut median decision latency by 22.4% versus DeepSeek and 61.9% versus Gemini.

TypeSafe’s Doom demo ran Jev at 10 queries per second. The team estimated the cost at roughly $7 per hour.

The table below covers Jev, its closest commercial rival, and the leading open-weight options. All specs come from each project’s own release page or model card.

These scores come from different test suites. Do not compare numbers across rows.

Two caveats matter here. Fastino chose both the tests and the opponents for its GLiDE and GLiNER2.5-Decide claims. Its Fast Decisions comparison also uses JevK5, an open reproduction, not TypeSafe’s Jev.

The idea is not new. Classifiers and rerankers have made decisions for years. Decision Transformer：https://arxiv.org/abs/2106.01345 (2021) framed reinforcement learning as sequence modeling. DeepMind’s Gato：https://arxiv.org/abs/2205.06175 (2022) showed one generalist model acting across many tasks. A 2023 survey：https://arxiv.org/pdf/2306.13945 even called ‘large decision models’ the next step.

What changed in 2026 is the packaging. 4 forces drove the shift:

Datalab Introduces OmniExtractBench to Fix Bias and Opacity in Extraction Benchmarks ：https://pxllnk.co/hxplrq

Practitioner-first AI/ML news and analysis, read by 1M+ developers and researchers every month.