Inkling

Inkling Open weights, ready to tinker Efficient open models for text, images, and audio, available to fine-tune. Two sizes of Inkling Inkling Our most capable open model, with native reasoning over text, images, and audio. Parameters975B total 41B activeArchitectureMixture of ExpertsContext1M tokens (64K / 256K on Tinker)InputText, Image, Audio Matches its larger sibling on many benchmarks

Inkling Open weights, ready to tinker Efficient open models for text, images, and audio, available to fine-tune. Two sizes of Inkling Inkling Our most capable open model, with native reasoning over text, images, and audio. Parameters975B total 41B activeArchitectureMixture of ExpertsContext1M tokens (64K / 256K on Tinker)InputText, Image, Audio Matches its larger sibling on many benchmarks at a quarter of the size, with lower cost and latency. Parameters276B total 12B activeArchitectureMixture of ExpertsContext1M tokens (64K / 256K on Tinker)InputText, Image, Audio Well-rounded,…

related reading

• Inkling: Our Open-Weights Model - Thinking Machines Labthinkingmachines.ai
• Welcome Inkling by Thinking Machineshuggingface.co
• What Thinking Machines’ Inkling Is Really Like, Part Isemaphore.substack.com
• A Safe Path to Open Weights - Thinking Machines Labthinkingmachines.ai
• This $12 billion startup finally shipped something...youtube.com
• Compare AI Models: Pricing, Context & Benchmarks | OpenRouteropenrouter.ai
• Tinkerthinkingmachines.ai
• Together AI | The AI Native Cloudtogether.ai
• Anatomy of a Modern Finetuning APIbenanderson.work
• Just get to the point.aimatey.co
• GitHub - JustVugg/colibri: Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦github.com
• MAI-Thinking-1: Building a Hill-Climbing Machinemicrosoft.ai