Thinking Machines Opens 975 Billion Parameter Model, Concedes It Trails Rivals
Inkling activates 41 billion of its 975 billion parameters per token, trained on 45 trillion tokens of text, image, audio and video.
- Post-training relied on large-scale asynchronous reinforcement learning across more than 30 million rollouts.
- Full weights are free on Hugging Face; fine-tuning runs on Tinker at 50% off.
- A smaller sibling, Inkling-Small, totals 276 billion parameters with 12 billion active.
Why it matters: Giving away a 975-billion-parameter model costs less than losing control of who trains the next one.
Thinking Machines Lab — Inkling launch post ↗ · Jul 15, 20267/15/26 · ✓ Checked✓ Check