Inkling: 1T Open-Source Multimodal Model
SonglinYang4 · x · 2026-07-16
vLLM shared an introduction to Thinking Machines' Inkling: a 1T parameter open-weights model supporting native multimodal input across text, image, and audio.
Key Details
- Maximum context length of up to 1 million tokens
- New architecture: relative attention, short convolutions, MoE expert sinks
- Equipped with 8 MTP heads for speculative decoding
- Supports NVFP4 and BF16 checkpoints
- Optimized for NVIDIA Blackwell and Hopper
- Achieves 380 tok/s/user on a 4× GB200 setup when combined with MTP
The official announcement confirms that full weights are open, available for fine-tuning on Tinker, and ready to test in the Inkling Playground.
More from Multimodal
- Apple's immersive horror short Remnants wows Vision Pro users with 3D presence — Scobleizer · 2026-10-18
- H3 360° Orbit Camera LoRA for MiniMax Open-Sourced, Generates Smooth 5.2s Shots — merica420_69 · 2026-10-18
- 'Tesla in Pakistan' AI video made with Seedance 2.5 goes viral — SimplyAnnisa · 2026-10-18
- Borges's nine stories turned into walkable 3D worlds with AI model Fable 5.1 — emollick · 2026-10-18
- Midjourney cozy couple photo: full prompt shared for killing the AI plastic look — tisch_eins · 2026-10-18
- Higgsfield playbook: scene templates, paid creator network, $20-to-enterprise funnel — sujingshen · 2026-10-18