Jev versus text-generating models
Jev does not generate free-form text. A side-by-side video compares parallel decision-making with token-by-token text generation.
Kev-0.6B Jev-class decision model runs one forward pass on Apple Neural Engine
Kev-0.6B is a Jev-class decision model running on the Apple Neural Engine. One forward pass, no token generation, 7.18 ms per decision with p99 of 7.53 ms. Placement isn't asserted yet; it's measured three ways: CPU-only returns zeros, and Apple's xctrace Neural Engine instrument counts 303.
Kev-0.6B, a Jev-class decision model, running on the Apple Neural Engine. One forward pass, no token generation. 7.18 ms per decision, p99 7.53. Placement isn't asserted yet, it's measured three ways: CPU-only returns zeros, Apple's xctrace Neural Engine instrument counts 303
· 1 likesOpen on X
More like this
Pricing, evaluations, and the name Jev
The launch thread explains code-and-AI workflow evaluations, the input price announced at launch, and free output tokens.
Jev Model: Decision-Making Probability Generation, 200× Faster and 400× Cheaper Than LLMs
This model is built for decision-making rather than text generation, outputting probabilities for each option in parallel instead of token-by-token responses. The author shares a simplified comparison of normal LLMs versus this Jev model.
