A fix in llama.cpp version b11523 lets WebGPU handle the DUP op, preventing crashes when running mixed‑batch models.
NoteBy OossaPublished by Oossa: 1 min read
The open‑source llama.cpp library released version b11523 on Oct 9, 2026. The update fixes a crash that occurred when the library tried to use the DUP operation on WebGPU. Previously, DUP ran on the CPU while other work ran on WebGPU, leading to a buffer mismatch and a crash. DUP now follows the same code path as the CPY (copy) and CONT ops, so it stays on WebGPU.
WebGPU users can run llama.cpp models with mixed batches without the program crashing.
Models
The company’s new three‑agent Claude‑based reviewer found 73.43% of core‑claim errors, far above the previous best of 14.81%.
Models
Mistral's new trillion‑parameter model scores 38 on an independent index, but costs more per task than competing open models.
Models
OpenAI reported two influence campaigns that used AI‑generated content to plant false stories in real media and blocked the involved accounts.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.
