Thinking Machines Launches Inkling, 975 Billion-Parameter AI Model Using Chinese Tech
Thinking Machines Lab has released Inkling, its first general-purpose foundation model. The mixture-of-experts model features 975 billion total parameters, 41 billion active parameters per token, and natively processes text, images, and audio inputs across a 1 million token context window.
The company stated that Inkling’s architecture is built on China’s DeepSeek-V3 and was refined post-training using data generated by Beijing-based Moonshot AI’s Kimi K2.5. Full weights are publicly available, with pricing and fine-tuning support immediately live on the Tinker platform alongside Day 0 inference partnerships.
From the sources (25 posts)
@victorialinmlRT @thinkymachines: Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making…
@techcrunchThinking Machines amps up its bet against one-size-fits-all AI with its first open model, Inkling
@thinkymachinesToday, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available. Available today for fine-tuning on Tinker. Play with it in the Ink
@natolambertPretty detailed safety section of model card, I found this interesting and practical: "Across all areas, we concluded that Inkling did not present risk of material uplift beyond what's already available in the open-weight ecosystem."
@thinkymachinesInkling is the first in a family. We’ve included some details of Inkling-Small, a lighter-weight model trained on a similar recipe, with full weights to follow.
@thinkymachinesWe hope you enjoy Inkling, and as always we’re keen to see what you build.
@_alex_kirillov_RT @thinkymachines: Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making…
@eliebakouch@thinkymachines congrats!! 🚀
@synthwaveddThinking Machines have released their first model, Inkling
@thezachmuellerRT @thinkymachines: Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making…
@miramuratiOur first model, Inkling. Trained from scratch, weights are open, fine-tunable on Tinker today.
@unslothai@thinkymachines Congrats guys on the launch! Inkling is the first open 1T multimodal model that supports image, audio, and text. 🤯🔥 We uploaded Dynamic GGUFs for those who want to run Inkling locally:
@stephenrollerRT @thinkymachines: Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making…
@mtsliveSITUATION DETECTED: Thinking Machines has released its first AI model, Inkling, an open-weight multimodal model built to be fine-tuned. Inkling is a 975B parameter MoE model with 41B active, a 1M token context window, trained on 45T tokens
@zephyr_z9New model from Thinky Looking strong They just need 1 or 2 post-training reps, and we have another true frontier model lab
@woosuk_kRT @thinkymachines: Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making…
@nvidiaaiCongrats @thinkymachines on the new open model 🙌 Inkling was trained on NVIDIA GB300 NVL72 and the NVFP4 checkpoint is available today on @huggingface: Happy building!
@zephyr_z9🫡🫡🫡🫡🫡
@samsja19RT @thinkymachines: Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making…
@soumithchintalaExcited for our first general model Inkling -- open weights, 975B, natively multimodal (text, image, audio). Available on Tinker, HuggingFace and partners. It is yours to personalize and use openly. It is yours.
@lmsysorgInkling, @thinkymachines' first open model, dropped today: 975B total / 41B active MoE, up to 1M context, reasoning natively over text, images, and audio. Serving and RL support are already live: you can run and shape it on an open stack,
@xianbao_qianWow a new open source model from @thinkymachines - 975B total, 41B active - 1M context length - Multimodal input text / image / audio inputs - Native nvfp4 weights - Fast inference infra ready on day 0 - Native MTP support - Apache 2.0 li
@_kevinluRT @thinkymachines: Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making…
@chhilleeIt truly takes a village to release a model, perhaps especially an open weights model. Actually doing the entire process from scratch, from data to pretraining to posttraining to actual release, gives a lot of appreciation for anyone who do
@techmemeThinking Machines Lab debuts Inkling, an open-weight MoE model with 975B total and 41B active parameters, trained to be broad rather than optimized for one area (Thinking Machines Lab) (Visit Techmeme dot com for the link and full context!