Skip to main content
Back to News
Mira Murati's Thinking Machines Lab debuts first AI model Inkling.
Product
2 min read
US

Mira Murati's Thinking Machines Lab debuts first AI model Inkling.

The AMW Read

A new entrant from a top-tier founding team with an open-weight strategy meaningfully updates the foundation model landscape, though early-stage release and modest claims limit immediate segment impact.
NoveltySignificance
Foundation Models · Player Map

Mira Murati's Thinking Machines Lab debuts first AI model Inkling.

Mira Murati, the former CTO of OpenAI who briefly served as CEO during Sam Altman's 2023 ouster, has unveiled Thinking Machines Lab's first AI model, called Inkling. The open-weight model was trained from scratch, and the company is explicitly setting low expectations: it is not the most performant model available today, closed or open. The blog post frames Inkling as a foundation for future models, built for solid capabilities across the board rather than state-of-the-art in any single area.

Why it matters: Thinking Machines Lab enters the foundation model arena at a moment when the cost of catching up to frontier labs has never been higher. By releasing an open-weight model with deliberately modest performance claims, Murati is positioning the startup in the shadow of DeepSeek and Meta's Llama lineage — a strategic choice that echoes the acqui-licensing pattern seen in earlier open-weight entrants. The move also tests whether a founder's reputation from one top-tier lab can substitute for massive compute budgets in the current capital-compression environment. The open-weight release gives Thinking Machines Lab a distribution channel without needing hyperscaler partnerships, but it also means the model will be immediately compared against Llama-4 and Qwen-3.5-class offerings, raising the bar for the company's next release.

Hayden Field, a senior AI reporter, noted the low-key framing from the company. The decision to train from scratch — rather than fine-tune an existing open model — signals that Murati's team intends to build independent technical IP, but it also commits them to a long and expensive training arc. The coming quarters will reveal whether Thinking Machines Lab can graduate from a foundational open-weight release to a genuinely competitive frontier model, or whether it becomes another cautionary add to the skeptic memory of high-profile founder startups that failed to reach escape velocity.

#MiraMurati #ThinkingMachinesLab #Inkling #OpenWeight #FoundationModels #AIStartups

#Mira Murati#Thinking Machines Lab#Inkling#open-weight model#foundation model

How This Connects

Based on Foundation Models · Player Map

  1. 1h agoAnthropic and Blackstone on July 15 formally launched Ode with Anthropic, a $1.5 billion enterprise...Ode
  2. 17h agoMira Murati's Thinking Machines Lab debuts first AI model Inkling. · THIS ARTICLE
  3. 1d agoDeepSeek reportedly raising $1.5B at $71B valuation, planning 2027 IPODeepSeek
  4. 3d agoMicrosoft Build 2026 unveils agent platform, reasoning model MAI-Thinking-1, and quantum chip Majora...
  5. 5d agoSAP Acquires Prior Labs in €1B Bet on Tabular Foundation ModelsPrior Labs
  6. 5d agoOpenAI launches 'ChatGPT Work' enterprise agent, integrates Codex into super-app desktop platform

Related News

More news from Thinking Machines Lab

Stay updated with the latest news and announcements from Thinking Machines Lab.

View all Thinking Machines Lab news

Discover AI Startups

Explore 2,000+ AI companies with VC-grade analysis, funding data, and investment insights.

Explore Dashboard