Fetching from the wire…
01
02
03
04
05
06
07
08
Infra2026-08-27 · source-backed
The August 26 release adds Decode Context Parallel for Kimi-K3, fused FlashKDA decode and prefill kernels, combined all-gathers with a claimed 1.5-3x kernel-level speedup, an adaptive speculative token budget worth about 60% better DSpark TTFT, and optional shared-expert sharding saving about 17 GiB per GPU (GitHub). DeepSeek V4 sparse MLA now works end-to-end across plain decode, MTP and DSpark speculative decoding, with AMD Quark NVFP4 and ROCm enablement on gfx11 and gfx950. If you self-host frontier open weights, this is the release that makes the two newest Chinese flagships cheap to serve.
Each link below shares sources, entities, or timing with this story.
Kimi K3 competes with DeepSeek V4 / Shared entities / Same source domain / Earlier coverage
Linked by a graph relationship (Kimi K3 competes with DeepSeek V4); both cover DSpark, Kimi; reported by the same outlet (github.com).
DeepSeek released DeepSeek V4 / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (DeepSeek released DeepSeek V4); both cover Chinese, GitHub; reported by the same outlet (github.com).
Kimi built by Moonshot AI / Shared entities / Earlier coverage / Tension
Linked by a graph relationship (Kimi built by Moonshot AI); both cover Chinese, DeepSeek V4; earlier Chinese coverage from 2026-05-05.
DeepSeek released DeepSeek V4 / Shared entities / Same source domain / Earlier coverage
Linked by a graph relationship (DeepSeek released DeepSeek V4); both cover DSpark, GPU, ROCm; reported by the same outlet (github.com).
Kimi K3 competes with DeepSeek V4 / Shared entity: Chinese / Shared topic / Earlier coverage
Linked by a graph relationship (Kimi K3 competes with DeepSeek V4); both cover Chinese; overlapping topics (better, chinese, deepseek).
DeepSeek released DeepSeek V4 / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (DeepSeek released DeepSeek V4); both cover DSpark, GitHub; reported by the same outlet (github.com).
Kimi built by Moonshot AI / Shared entities / Earlier coverage
Linked by a graph relationship (Kimi built by Moonshot AI); both cover GitHub, Kimi; earlier GitHub coverage from 2026-06-16.
Linked by a graph relationship (Kimi built by Moonshot AI); both cover Chinese, DeepSeek V4; earlier Chinese coverage from 2026-06-14.