The Inco Decode blog index lists three recent first-party announcements: the September 3, 2026 launch of the Inco platform on Artificial Analysis (covering Kimi K3, MiniMax M3, GLM 5.3, and GLM 5.3 Flash); an August 28, 2026 post on day-0 support for GLM 5.3 alongside DFlash 2 and NVFP4 checkpoints for faster, more eff The index confirms the chronology of Inco's stack evolution across late summer 2026, framing DFlash 2 as the underlying speculative-decoding technique and the GLM 5.3 day-0 launch as an application of that technique combined with NVFP4 quantization. The reference to "Muse Glimmer" appears only in this index entry and i

On September 3, 2026, Inco AI announced the public-beta launch of the Inco inference platform, releasing high-speed endpoints for Kimi K3, MiniMax M3, GLM 5.3, and GLM 5.3 Flash. According to the post, each model leads its respective Artificial Analysis provider leaderboard on output speed as captured on September 8, 2 The post details four stack layers that Inco credits for the performance results: (1) fleet-scale serving with cache-aware routing and scheduling; (2) speculative decoding via DFlash and DFlash 2, parallel block-diffusion techniques the post attributes to the Inco team, with open-source checkpoints already available; (

The incoai Hugging Face organization page lists Inco AI's published open-source artifacts, including DFlash and DFlash 2 text-generation checkpoints (1B, 2B, and 3B sizes), as well as DFlash 2 drafters for Qwen3.8-27B and GLM-5.3-Flash, plus GGUF variants. The page also shows a 391B text-generation model entry. These p Recent activity on the organization page shows updates from a team member (zhijianliu) within the last several hours and days, including updates to incoai/Qwen3.8-27B-DFlash2 (about 3 hours ago at capture) and incoai/GLM-5.3-Flash-DFlash2 (18 days ago), alongside a GLM 5.3 Flash DFlash 2 drafter updated 20 days ago. Th