Unveiling the New Competitor: Thinking Machines Introduces Inkling

Thinking Machines Lab, established by ex-OpenAI CTO Mira Murati, has unveiled their pioneering AI model named Inkling. Released on July 15, 2026, this marks a significant milestone for the company as they enter the competitive market of open large language models. With open weights and compatibility with the Tinker platform for fine-tuning, Inkling sets itself apart by offering a context window extending up to 1 million tokens.

Emphasizing Openness and Customization

Thinking Machines aims to craft adaptable AI models tailored to specific user needs rather than providing one-size-fits-all solutions. The company’s mission is clear: create AI that amplifies human decision-making capabilities.
Inkling leverages a Mixture-of-Experts architecture with 975 billion parameters, activating only 41 billion during operation. Its features include:

• A context window accommodating up to 1 million tokens.
• Multimodal functionality supporting text, images, and audio.
• Training on a vast dataset comprising 45 trillion tokens spanning text, images, audio, and video.
• An adjustable “thinking” level according to task complexity.

The company also introduced a preliminary version called Inkling-Small—a compact model with 12 billion active parameters—showing competitive performance in tests while requiring significantly fewer computational resources.

The Impact on the Market and Community Response

Inkling’s release has intensified competition among open-source large language models. Although not claiming to be the most powerful model available, its blend of multimodality and efficient processing makes it an excellent foundation for customization.
In testing scenarios like programming challenges and logical reasoning tasks, Inkling performs competitively. It achieved remarkable scores in various benchmarks such as AIME 2026 (97.1%), GPQA Diamond (87.2%), and SWE-Bench Verified (77.6%).
Importantly, Thinking Machines emphasizes resource efficiency; internal tests show that Inkling rivals Nemotron 3 Ultra in Terminal Bench using approximately one-third of the tokens.
Community feedback highlights Inkling’s unique combination of features rarely found together in an open solution: extensive token context windowing capability alongside multimodal support—all while maintaining high performance across diverse tasks including sound analysis and image processing.

Broader Implications for the Crypto Market

The launch of Inkling signifies a pivotal moment within AI development circles given its comprehensive feature set aimed at fostering innovation through openness coupled with customizable options—traits highly valued in crypto communities seeking flexible yet robust technological solutions adaptable across varying applications from blockchain analysis tools right down into decentralized finance platforms themselves.
As Mira Murati continues pushing boundaries beyond her previous tenure leading OpenAI initiatives toward even greater heights via Thinking Machine Labs’ endeavors now embodied chiefly through this latest creation known simply yet profoundly called “InkLing,” it remains evident how integral advancements like these shall undeniably shape future landscapes defining interactions between artificial intelligence paradigms coalescing seamlessly into ever-evolving cryptocurrency ecosystems worldwide today!