[OpenAI's Former CTO's New Company Open Sources First Model Inkling, Architecture Inspired by DeepSeek-V3] According to monitoring by Beating, OpenAI's former CTO Mira Murati's newly founded Thinking Machines Lab has released its first general-purpose model, Inkling. The model has a total of 975 billion parameters, with 41 billion activated per instance, and supports text, image, and audio inputs, with a maximum context length of 1 million tokens. Inkling was trained from scratch, but its MoE architecture primarily draws inspiration from DeepSeek-V3. During the early post-training phase, it also utilized data generated by open models such as Kimi K2.5. The complete weights have been uploaded to Hugging Face under the Apache 2.0 license. Thinking Machines explicitly acknowledges that Inkling is not currently the strongest model. Its focus is on customization and efficiency. On Terminal Bench 2.1, it achieves the same performance while using approximately one-third of the tokens required by NVIDIA's Nemotron 3 Ultra. [Original Link]