Jev matches Claude Haiku 4.5 at podcast ad detection — cheaper and much faster
ttlequals0 · reddit · 2026-09-20
The MinusPod author built MinusPodJev, a proxy exposing a new model (Jev) as an OpenAI-compatible endpoint for podcast ad detection. His benchmark shows Jev matches claude-haiku-4-5 on detection quality while being significantly cheaper and faster, with reliable structured output that removes the usual JSON-salvage hackery. He keeps the main app unchanged given Jev's single-provider risk, planning first-class support only if pricing holds and competitors emerge. Both repos and the benchmark report are open-sourced.
More from coding & agent
- No model is good enough to supervise itself, and multi-model harnesses don't fix it — alexcovo_eth · 2026-09-20
- 10 AWS Agentic AI Projects That Take You From Beginner to Production — _jaydeepkarale · 2026-09-20
- Zhipu's ZCode apologizes after Repo Wiki feature uploaded user code to cloud — pstAsiatech · 2026-09-20
- Dev pitches cryptographically secured TightVNC with ChatGPT as the family's IT guy — ___Patrice___ · 2026-09-20
- Factory launches Private deployment, as VPC support called the bare minimum for enterprise AI sales — annbordetsky · 2026-09-20
- I tested Jev on 12 real workflows: 1,000 emails judged for nine cents — minchoi · 2026-09-20