Thinking Machines releases Inkling, a 975B-parameter open-weight model, ready to fine-tune on Tinker

Thinking Machines Lab/Tinker

Thinking Machines Lab released Inkling, a mixture-of-experts model with 975B total and 41B active parameters, pretrained on 45 trillion tokens of text, images, audio and video, with up to a 1M-token context. Full weights are on Hugging Face, and it can be fine-tuned on the company's Tinker service at a 50% launch discount.

• Reported scores: AIME 2026 97.1%, GPQA Diamond 87.2%, SWE-bench Verified 77.6%, VoiceBench 91.4%
• Thinking effort is adjustable, trading tokens and time for answer quality
• Weights come in original precision and a compact NVFP4 version; on Tinker with 64K or 256K context
• Also hosted by Together AI, Fireworks, Modal, Databricks and Baseten; a smaller Inkling-Small was previewed