LaunchAI ModelsJuly 15, 2026
Thinking Machines releases Inkling, 975B open-weights multimodal model

Inkling is a 975B-parameter (41B active) Apache-2.0 multimodal MoE trained on 45 trillion tokens, accepting text, image, and audio input with 1M context. It is available day 0 on Together AI, and Unsloth GGUF plus NVFP4 quantizations are on Hugging Face.
Featured · Mira Murati
15 sources
Welcome Inkling by Thinking Machineshuggingface.co
Together AI brings Thinking Machines Lab’s new model Inkling on day 0together.ai
RT @mervenoyann: Inkling @thinkymachines 1-bit quant by @UnslothAI running 30-40 TPS 🤯 video sped...x.com
Inkling: Our open-weights modelsimonwillison.net
Inkling: A New Open-Weight 975B MoE with a Few Surprisessebastianraschka.com
[AINews] Thinky's Inkling: 975B-A41B multimodal, new best American Apache 2.0 open model (with Inkling-Small, 276B-A12B)latent.space
Murati’s Thinking Machines Releases First AI Model for Broad Usebloomberg.com
The new Inkling model was somewhat inevitable once Tinker took off -- integration of post-training s...bsky.app
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Anthropic found its models were used to hack 3 organizations
- Nous Research integrates Hermes Agent with Block's Buzz workspace
- China's MiniMax and ByteDance Release Dueling AI Video Models
- Tesla China rolls out Doubao-powered voice assistant to four vehicle lines
- Advancing responsible AI across Europe