September 19, 2026, (Inside AI) — Diogo Almeida, a former OpenAI researcher who helped build ChatGPT, has launched a new AI model called Jev that costs roughly 100 times less than comparable large language models. The model, released on September 18 by his startup TypeSafe AI, returns probability scores rather than text, a design choice that eliminates hallucination and slashes operating costs for automation and classification tasks.
Jev's launch comes as the AI industry shifts away from a single-best-model mindset. Decision-makers now prioritize cost, latency, and accuracy for specific jobs. A 625x price gap separates the cheapest usable option at $0.04 per million output tokens from frontier models charging $25 per million tokens. Jev targets that gap directly.
Almeida left OpenAI frustrated that ChatGPT, despite its success, lacked practical utility for machine-to-machine communication. He founded TypeSafe AI two years ago to build structured, deterministic outputs computers can act on without human interpretation. Jev represents that vision.
Instead of generating language tokens, Jev produces calibrated confidence scores between 0 and 1. Ask "Is this customer angry?" and Jev returns 0.9, meaning a 90% probability of yes. This numerical output cannot hallucinate because it never invents words. It also charges by the billion input tokens rather than counting output tokens, which is how traditional LLMs bill.
Read: Why AI’s Next Chapter Belongs to CPUs: The Shift to Efficient Inference
Developers adopted Jev immediately after launch, overwhelming TypeSafe's infrastructure. Pranit Sharma, an engineer at Vercel, tested Jev against OpenAI's Luna model and achieved results 5 to 18 times faster while maintaining better accuracy. Nikhil Mudholkar, CTO of Bryo AI, compared Jev to Gemini for classifying business emails. He found Jev cost 10 to 20 times less and returned actual probability scores ideal for automating workflows.
TypeSafe trained Jev exclusively on synthetic data. Almeida described this as his best strategic bet ever made. Synthetic data scaling enables human judgment at scale by automating large portions of annotation and data generation work. That approach avoids the cost and privacy issues of scraping real user data.
The broader AI landscape shifted dramatically throughout 2026 as specialized models began competing successfully against generalist language models. Enterprises face significant hidden costs after initial AI deployment. RaftLabs reports custom AI automation costs $30,000 to $120,000 for 3-10 workflows plus $500 to $8,000 monthly operational expenses. Jev's low API pricing directly attacks those recurring costs.
Pakistani startups should evaluate Jev carefully as a strategic component in their AI automation stacks. Pakistan's cost-sensitive technology market makes Jev particularly attractive compared to expensive frontier models. Reduced API expenses improve unit economics for startups building intelligent systems without access to VC funding subsidies, positioning them favorably against regional competitors.
TypeSafe AI has not disclosed total funding or customer count. The company also has not published independent benchmarks beyond user reports. Inside AI could not independently verify the performance claims made by Sharma and Mudholkar. However, the immediate infrastructure overload suggests genuine demand.
Jev's arrival signals a broader maturation in AI procurement. Buyers increasingly ask which model fits their specific cost, latency, and accuracy requirements rather than which model is universally best. That shift favors specialized tools like Jev over generalist chatbots. For automation and classification workloads, probability scores may prove more valuable than fluent prose. Almeida's bet on synthetic data and numerical outputs could redefine how enterprises measure AI return on investment.
