Thinking Machines Lab released Inkling on July 15—a 975-billion-parameter multimodal AI model with full weights available under an Apache 2.0 license on Hugging Face. The company was founded by Mira Murati, former OpenAI CTO who left in September 2024. Inkling scores 74.1% on MCP Atlas—nearly 30 points above Nvidia's Nemotron 3 Ultra—making it the best-performing Western open-weights model on agentic tool use, though Chinese models GLM 5.2 and Kimi K2.6 lead on several key benchmarks.
Inkling Model Specifications and Training Data
Inkling uses a mixture-of-experts architecture with 975 billion total parameters and 41 billion active per task. The model was trained from scratch on 45 trillion tokens spanning text, images, audio, and video. It accepts text, images, and audio inputs and supports a context window of 1 million tokens, roughly 750,000 words. Full weights are available on Hugging Face under an Apache 2.0 license with no restrictions. Fine-tuning is available through Thinking Machines' cloud platform Tinker.
Benchmark Performance Against Nvidia and Chinese Models
On MCP Atlas—which measures how reliably an AI agent completes real-world tasks using Model Context Protocol—Inkling posts 74.1%, nearly 30 points above Nvidia's Nemotron 3 Ultra. On SWE-Bench Verified—a test of whether an AI agent can autonomously fix real GitHub software bugs—Inkling scores 77.6%, above Nemotron's 70.7%. On FORTRESS Adversarial—which tests how consistently a model refuses genuinely harmful prompts without over-blocking legitimate ones—Inkling scores 78.0%, the highest mark among all open-weights models in the comparison.
Chinese models maintain leads on several fronts. Z.ai's GLM 5.2 scores 82.7% on Terminal Bench 2.1—a benchmark measuring autonomous AI coding agents in a real terminal environment—against Inkling's 63.8%. Kimi K2.6 leads on Humanity's Last Exam, a test of PhD-level scientific reasoning. Thinking Machines acknowledges Inkling is not the strongest model available today, open or closed, but positions it as the most capable open-weights model built by a Western lab.
Thinking Machines Lab Funding and Valuation History
Murati left OpenAI in September 2024 after serving as interim CEO for five days in November 2023 during Sam Altman's temporary removal. She founded Thinking Machines Lab in February 2025. The company raised $2 billion at a $12 billion valuation in July 2025, led by Andreessen Horowitz with Nvidia, Accel, ServiceNow, Cisco, AMD, and Jane Street participating—one of the largest seed rounds in Silicon Valley history at the time. Reports in November 2025 had the company seeking a new round at a $50 billion valuation. Those talks collapsed by January 2026.
Inkling-Small Model Preview
Thinking Machines previewed Inkling-Small: 276 billion total parameters, 12 billion active, matching the larger model on most reasoning benchmarks. The company stated its weights will arrive once testing is complete, with no timeline given.
FAQ
What did Thinking Machines Lab release on July 15?
Thinking Machines Lab released Inkling, a 975-billion-parameter multimodal AI model trained from scratch, with full weights available on Hugging Face under an Apache 2.0 license.
How does Inkling perform compared to Nvidia's Nemotron 3 Ultra?
Inkling scores 74.1% on MCP Atlas, nearly 30 points above Nvidia's Nemotron 3 Ultra, and scores 77.6% on SWE-Bench Verified, above Nemotron's 70.7%.
When did Mira Murati found Thinking Machines Lab?
Mira Murati founded Thinking Machines Lab in February 2025 after leaving OpenAI in September 2024.
