GLM-Omni Can Draw What It Cannot Say
Trained on 8,552 fictional images, Z.ai's in-house model drew a jersey number correctly 72% of the time and answered it in words 2% of the time. The paper's fix: let it draw first.
Read updateTrained on 8,552 fictional images, Z.ai's in-house model drew a jersey number correctly 72% of the time and answered it in words 2% of the time. The paper's fix: let it draw first.
Read updateAs a next-generation Flash model, M3.1 Flash delivers a generational leap in agentic tasks, coding, Computer Use and cybersecurity, while making online inference nearly three times faster and reducing usage costs by 70%
Read updateDeepSeek's V4.1 Flash 1017 announcement in full English: a stronger, cheaper Flash checkpoint, whole projects built from a single request, and search opening to the public.
Read updateMistral's first time-series foundation models top GIFT-Eval, TIME and fev-bench, and run on a CPU
Read updateIt out-scores Opus 5.5 on long-horizon coding by Meta's numbers, but trails it on the independent index
Read updateMistral's new 4.4B streaming speech-to-text model targets Arabic dialects and mid-sentence language switching
Read update