88Z AI, 13 November — Thinking Machines Lab has officially open-sourced the weights for its Inkling 2 Small model on Hugging Face under an Apache 2.0 license. Led by former OpenAI CTO Mira Murati, the 594-billion-parameter model supports ultra-long context and multimodal processing. This release marks the transition from closed testing to full openness, allowing developers to download and integrate the model immediately.
At a glance
- Inkling 2 Small weights are now open-source on Hugging Face under Apache 2.0
- The model features 594B parameters, 1M token context, and multimodal capabilities
- vLLM support is available, enabling immediate developer integration

Model Officially Open-Sourced
Thinking Machines Lab announced that the weights for Inkling 2 Small are now available for download on Hugging Face under the permissive Apache 2.0 license. Co-founder and former OpenAI CTO Mira Murati confirmed that this release fulfills their prior commitment, made possible by the team's efforts and collaboration with safety testing partners.
Technical Specifications and Performance
Inkling 2 Small is a 594-billion-parameter sparse mixture-of-experts (MoE) model with multimodal capabilities for handling both text and images, supporting a context window of up to 1 million tokens. The model demonstrates strong performance in coding, vision analysis, finance, and long-horizon reasoning, scoring 31 on the Artificial Analysis Intelligence Index. vLLM has rapidly added support, with reported inference speeds reaching up to 630 tokens per second.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- @thinkymachinesSource published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.