Thinking Machines Lab Drops Its First Model

2026-07-15 · Source: WIRED - Ai · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Fundamental Awareness, quick

Summary

Thinking Machines Lab has released Inkling, a 975-billion-parameter open-source model designed to understand video, audio, and text inputs. Trained from scratch, Inkling demonstrates strong performance in advanced reasoning and coding tasks, despite not leading popular benchmarks. The model notably used itself for fine-tuning, exhibiting a more concise "chain of thought" during efficient operation. This release aims to position Thinking Machines, founded in February 2025 by former OpenAI executives Mira Murati, John Schulman, and Lilian Weng, as a significant player in the AI landscape. The startup, which secured a \$12 billion seed funding round, advocates for decentralized AI and offers Inkling as a high-performance alternative to existing open-weight models, including those from China.

Key takeaway

For AI/ML directors evaluating new open-source foundation models, Inkling presents a compelling option from a well-funded startup with significant OpenAI talent. Its multimodal capabilities and reported strong performance in reasoning and coding suggest it could be a viable alternative to existing closed or open-weight models, potentially reducing operational costs and increasing customization flexibility. Consider exploring Inkling for your next project requiring advanced multimodal understanding and self-optimization potential.

Key insights

Thinking Machines Lab launched Inkling, a large open-source multimodal model, challenging established AI players with a vision for decentralized AI.

Principles

• Open-source models offer cost efficiency and modifiability.
• AI models can be used to fine-tune and improve themselves.
• AI technology should be decentralized, not controlled by a few companies.

Method

Inkling was trained from scratch on audio, video, and text inputs, then used its own capabilities for self-fine-tuning to enhance efficiency.

In practice

• Evaluate Inkling for advanced reasoning and coding tasks.
• Explore self-fine-tuning techniques for model optimization.

Topics

• Open-source AI
• Multimodal models
• Inkling
• Thinking Machines Lab
• Large Language Models
• AI decentralization

Best for: Investor, CTO, VP of Engineering/Data, AI Scientist, Director of AI/ML, Tech Journalist

Related on AIssential

• not much happened today — Inkling's launch emphasizes an open-weight, customizable multimodal foundation model designed for practical application and ecosystem integration over pure benchmark leadership.
• Together AI brings Thinking Machines Lab’s new model Inkling on day 0 — Inkling unifies multimodal inputs and offers controllable reasoning via a novel MoE architecture for broad task versatility.
• [AINews] Thinky's Inkling: 975B-A41B multimodal, new best American Apache 2.0 open model (with Inkling-Small, 276B-A12B) — Inkling is an Apache 2.0 multimodal MoE model, prioritizing customization and efficiency over raw benchmark supremacy.
• Thinking Machines Lab’s First Model Gives Nod to Chinese Open Source AI — Thinking Machines Lab's Inkling model demonstrates leveraging established open-source architectures, specifically DeepSeek V3, for new AI development.
• Thinking Machines open sources first multimodal language model, Inkling, focused on low cost and 'resistance to censorship' — Inkling offers a natively multimodal, open-source foundation model with programmatic cost-performance control and resistance to censorship.
• Inkling: Our open-weights model — Inkling is an Apache-2.0 licensed multimodal base model for fine-tuning, not a frontier model.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by WIRED - Ai.