SGLang v0.5.16 adds DSpark, cuts GLM-5.2 KV memory and supports Inkling on day one

BanghuaZ · x · 2026-07-25

SGLang v0.5.16 ships a new speculative decoding algorithm called DSpark and several inference/runtime optimizations.

• DSpark is reported to stay ahead of MTP across concurrency levels, with 383.7 tok/s at accept length 5 on DeepSeek-V4-Pro (TP8, B300, bs=1).
• Day-0 support is added for Inkling, the first open-weights model from Thinking Machines Lab, with verified recipes for NVIDIA Blackwell, H200, and AMD MI350X/MI355X.
• Other newly supported models include LongCat 2.0 FP8, JetBrains Mellum v2, Pi0.5, and LongLive 2.0.
• The release also introduces one shared prefix cache for SWA/Mamba/DSA models, GPU cache splitting for GLM-5.2, lighter speculative decoding for GDN, and the first correct KDA MTP path on Blackwell.

Original post →

More from Infra

• AI chip stacks are still wide open, with memory, networking, and optics left to improve — bookwormengr · 2026-07-25
• NVIDIA says it is partnering with Korea on chips, robotics and AI factories — nvidia · 2026-07-25
• Huawei’s 950 NPU reportedly uses 4 dies and separate SerDes dies — teortaxesTex · 2026-07-25
• ParseBench puts Claude Opus 5 near Opus 4.8 on docs, but cheaper rivals win on tables — llama_index · 2026-07-25
• Databricks says Genie Code beat three coding agents on 401 real tasks at $0.55 each — DbrxMosaicAI · 2026-07-25
• Alphabet’s future spending commitments jump to $811B as Stripe reportedly weighs OpenRouter — 创业邦 · 2026-07-25

Infra channel →