Jev 1.14 can see, and Clef claims 3 of 4 against it
TypeSafe says Jev was an image model the whole time and they had it switched off. Same day, Merge says Cloudflare's own tests put Clef ahead of Jev on three of four workflow evals.
Built with it
Tony Dinh dropped a $124 a month i18n SaaS for a $0.75 run
Tony Dinh cancelled the translation SaaS he was paying $124 a month for and moved i18n into a built-in translation harness. He ran 2,705 translation keys across 16 locales for about $0.75 with GPT-Luna 6. The harness feeds the model app context: internal words, which feature and screen a string appears on, and how much space it has.
Someone measured it
Jev 1.14 adds images, a box type and 262k context at the same price
TypeSafe shipped Jev 1.14 with image input, a bounding box type and context up to 262k, all behind opt-in danger zone headers. OpenRouter lists it at an unchanged $0.042 per million input tokens, output free, and calls multilingual the biggest upgrade. Diogo Almeida says the model was multimodal all along and they had disabled it on purpose.
Yep, the Jev you know and love was an image model this whole time.
The case against
Clef beat Jev on 3 of 4 workflow evals, by Cloudflare's count
Merge put Cloudflare's Clef models on its gateway and says that in Cloudflare's tests Clef came out ahead of Jev on three of four workflow evals. It also puts Clef-flash at a median 39ms. The tasks, the eval set and which Jev version was tested are not described anywhere public.
Clef-flash answers in a median 39ms, and in Cloudflare's tests Clef came out ahead of Jev on 3 of 4 workflow evals.