Crypto
Home›Crypto›Market Structure›Thinking Machines Lab launches Inkling, a 975B open-so…
Thinking Machines Lab launches Inkling, a 975B open-source model
The model is offered via OpenRouter at $1 per million input tokens and $4.05 per million output tokens, and it posts an MCP Atlas score of 74.1%.
Thinking Machines Lab has released Inkling, a 975 billion-parameter open-source AI model, making its debut on July 15. Decrypt reports it is Mira Murati’s first major model release since she left OpenAI in September 2024 and is the lab’s first model since a two-year silence.
Inkling is live on OpenRouter with pricing set at $1 per million input tokens and $4.05 per million output tokens. The release is distributed with full weights on Hugging Face under the Apache 2.0 license and is described as a mixture-of-experts design with 975 billion total parameters and 41 billion active at inference.
The model supports a 1 million token context window and was pretrained on 45 trillion tokens, according to the article. It can read text, images, and audio, and it is positioned as having a clear advantage in agentic tool use through the Model Context Protocol standard.
In benchmarks shared in the article, Inkling’s MCP Atlas score is 74.1%, nearly 30 points above Nvidia’s Nemotron 3 Ultra, and it scores 77.6% on SWE-Bench Verified for autonomous GitHub bug fixing, above Nemotron’s 70.7%. Decrypt also says the model is presented as fully private, while its hands-on performance is described as holding up but disappointing in parts.