Fetching from the wire…
01
02
03
04
05
06
07
08
Infra2026-08-09 · source-backed
DCP splits KV cache across GPUs by token position rather than attention head, eliminating the duplication that caps tensor parallelism on long contexts (vLLM). On 8×B200 with Kimi K2.6, baseline TP plateaued at ~1,863 tok/s/GPU at concurrency 64 while DCP reached 6,091 tok/s/GPU at concurrency 512 using only 82% of KV memory, staying stable past 200K-token sequences. Enable with --tensor-parallel-size N --decode-context-parallel-size M. Helps MLA/GQA models on 64K+ contexts with large-input/short-output agentic traffic, which describes basically every coding agent. Hurts on short contexts or weak GPU interconnect.
Each link below shares sources, entities, or timing with this story.
Moonshot AI deprecates Kimi K2 / Shared entity: GPUs / Shared topic / Earlier coverage
Linked by a graph relationship (Moonshot AI deprecates Kimi K2); both cover GPUs; overlapping topics (agentic, coding).
Moonshot AI deprecates Kimi K2 / Shared entity: Kimi K2 / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Moonshot AI deprecates Kimi K2); both cover Kimi K2; overlapping topics (agent, agentic, coding).
K2 Vendor Verifier benchmarked against Kimi K2 / Shared entity: Kimi K2 / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (K2 Vendor Verifier benchmarked against Kimi K2); both cover Kimi K2; overlapping topics (agent, agentic, coding).
Kimi K3 benchmarked against B200 / Shared entity: GPU / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Kimi K3 benchmarked against B200); both cover GPU; overlapping topics (b200, decode).
Moonshot AI deprecates Kimi K2 / Shared entities / Earlier coverage
Linked by a graph relationship (Moonshot AI deprecates Kimi K2); both cover GPU, Kimi K2; earlier GPU coverage from 2026-03-23.
Shared entities / Shared topic / Earlier coverage / Downstream implication
Both cover GPU, GPUs; overlapping topics (context, parallelism); earlier GPU coverage from 2026-04-10.
Moonshot AI deprecates Kimi K2 / Shared entity: Kimi K2 / Shared topic / Earlier coverage
Linked by a graph relationship (Moonshot AI deprecates Kimi K2); both cover Kimi K2; overlapping topics (agentic, coding).
Linked by a graph relationship (Moonshot AI deprecates Kimi K2); both cover Kimi K2; overlapping topics (agent, coding).