https://derekwang2002.github.io/ai-intelligence/en/projects/jev-35b/ Material changes in AI models, agents, engineering and infrastructure, with primary sources and recommendations. en Mon, 02 Nov 2026 16:56:00 GMT https://derekwang2002.github.io/ai-intelligence/en/events/ev-20261011-03/ ev-20261011-03-published Tue, 13 Oct 2026 16:05:06 GMT New IGNORE <p>AutoTrust AI released Apache-2.0 weights for JEV-35B, a Qwen3.5-35B-A3B fine-tune for ranking up to 256 options in one pass. The model card reports a self-scored public benchmark and a 204 ms median on one B200. The full benchmark is mostly private, the result is not a board entry and no independent deployment evidence is available.</p><p>Recommendation: IGNORE</p><p>Sources: <a href="https://huggingface.co/autotrust/JEV-35B">JEV-35B model card</a></p>