TypeSafe AI has launched Jev, a new classification model designed for software automation, which is 100 times cheaper than traditional large language models (LLMs). The model has already gained significant traction among developers, outperforming OpenAI's Luna in both speed and cost.
TL;DR
- TypeSafe AI releases Jev, a classification model optimized for automation tasks, costing 100x less than LLMs.
- Jev outperforms OpenAI's Luna in speed and cost, with developers reporting 5 to 18 times faster results.
- The model's ability to provide calibrated decisions and probability scores makes it ideal for high-volume enterprise tasks.
What happened
TypeSafe AI, founded by Diogo Almeida, a former OpenAI engineer who contributed to ChatGPT, has launched Jev. The model is designed specifically for automation workflows and classification tasks, providing probability scores and calibrated decisions instead of generating text.
Jev charges by the billion input tokens, making it approximately 100 times cheaper than comparable models. This pricing structure makes it viable for high-volume tasks at enterprise scale.
Developers have quickly adopted Jev, with Vercel engineer Pranit Sharma reporting 5 to 18 times faster results compared to OpenAI's Luna model while maintaining better accuracy. Bryo AI CTO Nikhil Mudholkar found Jev to be 10 to 20 times cheaper than Gemini for classifying business emails, with actual probability scores ideal for automating workflows.
Why it matters
The launch of Jev signifies a shift in the AI landscape, where specialized models are beginning to compete successfully against generalist language models. This trend is driven by the need for models that fit specific cost, latency, and accuracy requirements.
For developers and startups, Jev offers a cost-effective solution for high-volume automation tasks. Its ability to provide calibrated decisions and probability scores makes it particularly useful for enterprise-scale applications.
Investors should take note of the growing market for specialized AI models. The success of Jev highlights the potential for significant cost savings and improved unit economics in AI automation stacks.
Key facts
- Jev is designed for automation workflows and classification tasks, providing probability scores and calibrated decisions.
- The model is 100 times cheaper than traditional LLMs, charging by the billion input tokens.
- Jev cannot hallucinate as it outputs numerical probabilities rather than language tokens.
- Developers have reported 5 to 18 times faster results compared to OpenAI's Luna model with better accuracy.
- Jev is 10 to 20 times cheaper than Gemini for classifying business emails, according to Bryo AI CTO Nikhil Mudholkar.
- TypeSafe trained Jev exclusively on synthetic data, enabling human judgment at scale by automating large portions of annotation and data generation work.
- Enterprise customers face significant hidden costs after initial AI deployment, with custom AI automation costing $30,000 to $120,000 for 3-10 workflows plus $500 to $8,000 monthly operational expenses, according to RaftLabs.
- A remarkable 625x price gap exists between the cheapest usable option costing $0.04 per million output tokens and frontier models commanding $25 per million tokens.
Context
The AI landscape has seen a dramatic shift in 2026, with specialized models beginning to compete against generalist language models. This trend is driven by the need for models that fit specific cost, latency, and accuracy requirements.
TypeSafe AI's Jev model represents a significant advancement in this trend, offering a cost-effective solution for high-volume automation tasks. Its ability to provide calibrated decisions and probability scores makes it particularly useful for enterprise-scale applications.
The success of Jev highlights the potential for significant cost savings and improved unit economics in AI automation stacks, making it an attractive option for cost-sensitive technology markets like Pakistan.