Fetching from the wire…
Public story · 2026-08-29 · source-backed
Build b10677 fixes ggml_vk_graph_optimize, where is_src_of didn't treat two views of one tensor as dependent, so the optimizer reordered nodes across aliased reads and writes. Maintainers describe the result as silently wrong tokens under greedy decoding, different output on every server start, and invalid speculative-decoding acceptance, hitting Qwen3.8's recurrent state and any model with view-aliased state. CUDA was unaffected. Nothing logged. Anyone benchmarking Vulkan inference on those models in the last stretch has results that need re-running. (GitHub)
Each link below shares sources, entities, or timing with this story.
NVIDIA released DGX Spark / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (NVIDIA released DGX Spark); both cover AMD, CUDA, NVIDIA, Vulkan; overlapping topics (cuda, token).
NVIDIA released Blackwell / Shared entities / Same source domain / Earlier coverage / Downstream implication
Linked by a graph relationship (NVIDIA released Blackwell); both cover AMD, CUDA, NVIDIA, Qwen3; reported by the same outlet (github.com).
NVIDIA invested in OpenAI / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (NVIDIA invested in OpenAI); both cover AMD, CUDA, NVIDIA, Qwen3; overlapping topics (model, token).
NVIDIA partners with Cisco / Shared entities / Earlier coverage
Linked by a graph relationship (NVIDIA partners with Cisco); both cover AMD, CUDA, GitHub, NVIDIA; earlier AMD coverage from 2026-07-27.
NVIDIA invested in OpenAI / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (NVIDIA invested in OpenAI); both cover Anyone, GitHub; reported by the same outlet (github.com).
NVIDIA uses Claude Code / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (NVIDIA uses Claude Code); both cover GitHub, Qwen3; reported by the same outlet (github.com).
NVIDIA released DGX Spark / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (NVIDIA released DGX Spark); both cover NVIDIA, Qwen3; reported by the same outlet (github.com).
NVIDIA uses Claude Code / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (NVIDIA uses Claude Code); both cover GitHub, Qwen3; overlapping topics (model, token).