Inkling Model Specs, Architecture Details & SGLang Support
BanghuaZ · x · 2026-07-16
Thinking Machines' open-source model Inkling boasts 975B total parameters and 41B active parameters (MoE architecture), supports up to a 1M context, and natively handles text, image, and audio reasoning.
Architecture & Inference Optimization:
- Employs a novel architecture, including ShortConv, relative positional encoding attention, and shared expert MoE.
- SGLang now offers comprehensive support: including PD disaggregation, multi-LoRA serving, HiCache, and a DFlash checkpoint trained by Modal.
- Achieves full CUDA graph optimization during the prefill phase and utilizes an MXFP8 KV cache.
More from Infra
- DeepSpeed shows AutoEP/AutoTP/AutoSP parallelism gains, hosts office hours — PyTorch · 2026-10-27
- Running a local 9B as a coding agent: brief the big model to plan, let the small model execute — Electrical_Ninja3805 · 2026-10-27
- Cloudflare Durable Objects validated as top architecture for AI agent WebSocket servers — tristanbob · 2026-10-27
- Gary Marcus: With Q3 data added, Anthropic's compute-efficiency growth shows deceleration — GaryMarcus · 2026-10-27
- SpaceX locks in $40B GPU financing deal with Apollo, the largest ever committed — TechCrunch AI · 2026-10-27
- Inference is just Kubernetes with a multi-trillion-dollar market, and the communities overlap — matt_slotnick · 2026-10-27