Thinking Machines Releases 975B-Parameter Inkling as US Open-Weight Alternative
Updated
Updated · InfoWorld · Jul 16
Thinking Machines Releases 975B-Parameter Inkling as US Open-Weight Alternative
3 articles · Updated · InfoWorld · Jul 16
Summary
Inkling debuted as Thinking Machines Lab’s first general-purpose model, giving Western enterprises a US-developed open-weight option in a market where Chinese models dominate several coding and reasoning categories.
975 billion total parameters—41 billion active—power a multimodal mixture-of-experts model with a 1 million-token context window, trained on 45 trillion tokens and exposed through Tinker, Hugging Face and major inference platforms.
77.6% on SWE-Bench Verified put Inkling behind DeepSeek V4 Pro and GLM 5.2 but ahead of Nvidia Nemotron 3 Ultra; Thinking Machines also said it matched Nemotron on Terminal Bench 2.1 using about one-third as many tokens.
2 TB of aggregated VRAM is needed to run the BF16 checkpoint privately—roughly eight Nvidia B300s or 16 H200s—while a quantized version cuts that to 600 GB, leaving cost a key hurdle for enterprise deployment.
276 billion parameters in the previewed Inkling-Small could broaden adoption, while analysts said enterprises will need post-fine-tuning safety retests and governance because customized self-hosted models can drift from the official release.
Can a business based on customizing open AI truly challenge the pay-per-use giants like OpenAI and Google?
As Europe seeks digital sovereignty, will this new open-weight AI become its weapon against American tech dominance?
From Zero to Market in 9 Months: The Launch of Inkling and Thinking Machines Lab’s Bid to Transform Enterprise AI
Overview
On July 15, 2026, Thinking Machines Lab officially launched Inkling, marking a pivotal moment in the AI industry. Inkling stands out for its remarkably rapid development, reaching market presence and generating revenue in just nine months—a pace that sharply contrasts with industry leaders like OpenAI and Anthropic, which took years longer. This swift entry highlights Thinking Machines Lab’s efficient execution and ambition, positioning Inkling as a dynamic new competitor. The company’s accelerated timeline and innovative approach signal a significant shift in how quickly advanced AI solutions can be brought to market and adopted by enterprises.