Thinking Machines Lab

Yuchen Jin

@Yuchenj_UW

We’re excited to announce that Databricks is a Day-0 launch partner of Thinking Machines Lab (@thinkymachines), bringing its first oss model, Inkling, to the Databricks platform. - the strongest US oss model - 974B total parameters, 41B active MoE - Apache 2.0 license The Thinky team is absolutely cracked. We’ve loved working with them. Come try Inkling on Databricks!

Rohan Paul

@rohanpaul_ai

Techcrunch: Lilian Weng is leaving Thinking Machines Lab for OpenAI, moving from co-founder role into a narrower research post. Weng said startup stress and workload had pushed her health beyond what she could sustain, after months of considering the decision. She also said a scoped role in a more predictable place, without co-founder responsibility, would suit her better. OpenAI says she will lead a top-level group focused on accelerating internal research into recursive self-improvement. Weng returns with direct institutional knowledge from 2018 to 2024, including GPT-4 development and leadership of Safety Systems. Her appointment puts a former safety leader near work that OpenAI’s Preparedness group is already hiring researchers to manage. For Thinking Machines Lab, the departure deepens a retention problem, with only 2 of the original 6 co-founders remaining. --- techcrunch .com/2026/07/29/thinking-machines-co-founder-lilian-weng-left-the-company-citing-health-reasons-then-joined-openai/

Chubby♨️

@kimmonismus

Oh and dont sleep on this: Mira Muratis Thinking Machines Lab just released Inkling-Small: a 276B MoE open weight model that activates only 12B parameters per token. It beats the 41B-active Inkling on Terminal-Bench 2.1 (64.7 vs. 63.8), HLE (31.6 vs. 29.7), and SWE-Bench Verified (80.2 vs. 77.6). The company says it matches or beats the much larger Inkling on reasoning and agentic coding, while offering native audio and vision, a 1M-token context window, and variable thinking effort. On-policy distillation plus two weeks of coding RL made the smaller model outperform its teacher on several benchmarks. Inkling-Small could become a remarkably efficient open foundation for multimodal agents.

Sriram Krishnan

@sriramk

It is clear open source models and harnesses are having a moment. There's a few factors at work 1/ It is now obvious that you can catch up to near-SOTA performance and do so with a clear training lineage. See:@thinkymachines Inkling launch today. 2/ There are several well-funded, talented teams building open weight models now in the US and abroad. Along with the explosing of other near SOTA models (Grok/Cursor, Muse Spark), it is clear we are going to have a diverse ecosystem of models atleast on coding and agentic use. 3/ Organizations are increasingly looking for control over how their data is used and are willing to trade off some access to frontier level tokens for this control. Organizations and countries are increasingly nervous about the frontier labs potentially competing with them down the road and don't want their data to enable a future competitor. 4/ Open source is a slider: you could bring your own open harness, your evals, your business context and are free to pick and choose your model of choice. 5/ Companies have now actively shifted from "how do we get our people to use tokens" to being uncomfortable with their token cost ballooning without a clear line to revenue. 6/ Geo-politically, countries will be weighing open weight models as a way to get frontier-level tokens inside controlled environments that may not be otherwise possible. All of this leads to more choice for all of us !

Artificial Analysis

@ArtificialAnlys

Thinking Machines' new Inkling Small scores 40 on the Artificial Analysis Intelligence Index, within a point of its flagship sibling Inkling with less than a third of the total and active parameters Inkling Small is @thinkymachines second model release, arriving two weeks after Inkling launched at 41 on the Artificial Analysis Intelligence Index. Thinking Machines is the San Francisco-based AI lab founded by former @OpenAI CTO Mira Murati. Inkling Small is an open weights reasoning model with 276B total parameters (12B active MoE), text, image, and speech input, and a 256K token context window. Key results: ➤ Inkling Small holds a similar tier of intelligence to Inkling’s at less than one third of its size: 276B total parameters (12B active) vs. 975B (41B active) for Inkling. No open weights model at its size or smaller scores higher on the Intelligence Index. DeepSeek V4 Flash (max), at a similar size of 284B total (13B active) also scores 40, MiniMax-M3 reaches 44 with 23B active, while GLM-5.2 (max) reaches 51 with 40B active. ➤ Inkling Small meets or exceeds Inkling on several coding and frontier reasoning evaluations. It scores higher on Humanity's Last Exam (32% vs. 30%), GPQA Diamond (89% vs. 87%), CritPt (8% vs. 5%), and SciCode (49% vs. 46%), and achieves the same score on Terminal Bench v2.1 (55%). ➤ Inkling Small is not as strong as Inkling on Agentic tasks and factual knowledge. Inkling Small trails Inkling on τ³-Banking (15% vs. 24%), though it edges ahead on GDPval-AA v2 (1269 vs. 1237 Elo). On the AA-Omniscience Index it scores -9 vs. the flagship's positive 2, meaning incorrect answers outweigh correct ones; this is driven by lower AA-Omniscience Accuracy (31% vs. 40%) rather than hallucination: Inkling Small’s Hallucination Rate is slightly lower than its larger sibling (57% vs. 63%). ➤ Inkling Small averaged ~24K output tokens per Intelligence Index task, slightly fewer than Inkling (~25K), while peers at its intelligence level averaged far more. DeepSeek V4 Flash averaged ~45K and GPT-5.4 mini (xhigh) ~78K. Running the full Intelligence Index took roughly the same number of output tokens for Inkling Small and Inkling. Additional model details: ➤ Type: Open weights reasoning model (Apache 2.0 license) ➤ Size: 276B total parameters, 12B active (MoE) ➤ Input modalities: Text, image, and speech ➤ Output modalities: Text ➤ Context window: 256K tokens (Inkling supports 1M) Congratulations to the team at Thinking Machines on the release!

Xiaoyin Qu

@quxiaoyin

My bet: @thinkymachines will soon make more money than @AnthropicAI. Not by winning the race to build one standardized frontier model. By becoming the Palantir FDE for enterprise custom models. The playbook: 1. Release the best American open-weight model. 2. Drive widespread enterprise adoption. 3. Charge the largest companies 7–9 figures to post-train and run custom models behind their own firewall. The model rests on three bets: 1. Large enterprises will increasingly demand their own models with their own data, and this is how they differentiate and win. 2. Enterprises won’t need just one model. They’ll continuously need new models for different workflows, departments, and proprietary datasets. That creates extremely sticky, recurring revenue. 3. Autoresearch will make custom model development increasingly scalable. Tinker can become the interface enterprises use to post-train their own models—with @thinkymachines providing the expertise and infrastructure behind it. FDE, infra, everything, huge contracts. 4. Eventually, maybe everyone wants their OWN model, and autoresearch and training inside tinker on top of @thinkymachines's base model will make it happen. Meanwhile, Henry-ford-styled, standardized models will makes no margins. OpenAI and Anthropic will have their API margins squeezed by Deepseek/GLM/Grok/Meta etc, and their consumer subscriptions are loss centers. The fat margin will move to customization: proprietary data, post-training, evals, deployment, and infrastructure. If this thesis is right, @thinkymachines isn’t building just another frontier lab. It’s building the highest-value layer between frontier research and enterprise model ownership. Turns out, the best business model for enterprise is NOT to sell commodity API access. Sell them their own models. I’m extremely bullish on this approach. @miramurati may be the most commercially savvy frontier-lab leader. I have to admit it.

Mira Murati

Co-founder & CEO

Former Chief Technology Officer of OpenAI; founded Thinking Machines Lab in 2025.

John Schulman

Co-founder & Chief Scientist

OpenAI co-founder and a key architect of ChatGPT; chief scientist at Thinking Machines Lab.

Lilian Weng

Co-founder

Former OpenAI VP of research and safety; co-founded Thinking Machines Lab in 2025.

Soumith Chintala

Chief Technology Officer

Creator of PyTorch and long‑time Meta AI researcher; appointed CTO of Thinking Machines Lab after Barret Zoph’s departure.

Anton Oyung

Member of Technical Staff

Previously at Roblox · San Francisco Bay Area

Joined Jun 2026

Paul Z.

Member of Technical Staff

Previously at Meta · New York, New York, United States

Joined Jun 2026

Jack Cai

Member of Technical Staff

Previously at xAI · San Francisco Bay Area

Joined May 2026

Previously at xAI · Mountain View, California, United States

Joined Jun 2026