Breaking

Thinking Machines Lab Releases Its First Open-Weight Model, Inkling, Betting on Customizable AI

July 15, 2026 at 14:07 EDT

Foundation Models
Open Source

Thinking Machines Lab, the startup founded by former OpenAI CTO Mira Murati, released its first open-weight model on July 15, 2026, called Inkling, doubling down on its argument that customizable models beat one-size-fits-all systems. The launch positions the company against rivals whose flagship products are closed, general-purpose chatbots.

July 15, 2026 · Thinking Machines Lab

Inkling: A Bet Against One-Size-Fits-All AI

Mira Murati's startup releases its first open-weight model — a multimodal, customizable foundation built to be fine-tuned, not to top leaderboards. The thesis: enterprises will prefer models they can own and specialize over rented, closed systems.

975B

total parameters (MoE)

~41B

active at inference

45T

training tokens · text, image, audio, video

Efficiency, drawn to scale

Matching Nvidia's Nemotron 3 Ultra on coding — using about a third of the tokens.

3× tokens

Nemotron 3 Ultra

1× tokens

Inkling

Proof point · Bridgewater case

An open-source model trained on financial expertise — the customization thesis in numbers. (Predates Inkling; not a direct measure of the new model.)

84.7%

on a financial-reasoning test

~1/14

operating cost vs a proprietary alternative

What sets Inkling apart

Open weights — download, modify and run it yourself, fine-tuned via the Tinker platform.
Native multimodal — text, images, audio and video reasoning; strong on VoiceBench, MMAU, AudioMC.
Adjustable "thinking effort" — tune how much compute goes into reasoning, with explicit uncertainty.
Well-rounded, not chart-topping — trained on Nvidia GB300 NVL72; a lighter Inkling-Small variant to follow.

Still early: launched only on release day, Inkling has no independent third-party benchmarks or production reports yet. Its success hinges on how developers adapt it through Tinker — and how Inkling-Small lands once its weights arrive.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.