What are the key specifications of Inkling?

Inkling is a mixture-of-experts model with 975 billion parameters, trained on about 45 trillion tokens of text, image, audio and video. For the average prompt, it draws on only about 41 billion parameters to process tasks faster and keep costs low. It can reason natively across all four input types but outputs text only, including code, styled artifacts and structured data.

How does Thinking Machines plan to make money?

Rather than charging customers for model access via API, Thinking Machines plans to generate revenue through Tinker, a paid service that makes it simple for developers to fine-tune open-weights models for specific tasks. The model itself is available to download with full open weights.

How does Inkling compare to proprietary models in cost and performance?

In early test results with Nvidia's Nemotron 3 Ultra model, Inkling achieved comparable coding performance despite using two-thirds fewer tokens. In a collaboration with Bridgewater Associates, a fine-tuned version scored 84.7% on leading financial reasoning benchmarks, outperforming the most advanced proprietary alternatives at less than 10% of the cost.

Large Language ModelsSiliconANGLE AIJul 16, 2026Jul 16, 2026, 16:01 JST3 min read

Mira Murati's Thinking Machines launches Inkling, open-weight AI model

Send on LINE

Key takeaway

Thinking Machines Lab Inc., founded by former OpenAI CTO Mira Murati, has launched Inkling, an open-weights AI model with 975 billion parameters trained on 45 trillion tokens of multimodal data.
Unlike proprietary rivals that charge per API token, Thinking Machines plans to monetize through Tinker, a paid platform for fine-tuning.
The release positions Inkling as a Western alternative to lower-cost Chinese open-source models and appeals to organizations seeking to customize AI on their own infrastructure rather than rely on expensive licensing.

3 Key Points

What happened
Thinking Machines Lab Inc., founded by former OpenAI CTO Mira Murati, released Inkling, its first foundation model trained from scratch. The model is a mixture-of-experts system with 975 billion parameters, trained on about 45 trillion tokens of text, image, audio and video, and is available as open weights that developers can download and fine-tune without licensing fees.
Why it matters
Inkling fills a gap in the Western open-source AI ecosystem, which has lagged behind China's, particularly as Meta downplayed its Llama models in favor of proprietary systems. The open-weights release gives Western enterprises an alternative to lower-cost Chinese AI models and lets developers customize the model for their own infrastructure rather than paying per-token API fees—a shift that could reshape how organizations evaluate and deploy AI.
What to watch
Thinking Machines plans to generate revenue through Tinker, its paid fine-tuning platform (launched in October), rather than charging for model access. In a test with Bridgewater Associates, researchers fine-tuned an open model with financial data and achieved 84.7% on leading financial reasoning benchmarks, outperforming proprietary alternatives at less than 10% of the cost. The company built Inkling from scratch in less than nine months using Nvidia's GB300 NVL72 system.

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Get news like this every morning — free

Context & Analysis

Thinking Machines' launch of Inkling marks a strategic pivot in the open-source AI landscape. The company, which spent the past year primarily announcing funding rounds and its Nvidia partnership, now enters the market with a fully homegrown foundation model trained in under nine months—substantially faster than the multiyear timelines at rivals like OpenAI and Anthropic. The model's open-weights availability directly counters a year-long dominance by Chinese AI firms in the open-source ecosystem, a gap that widened after Meta deprioritized its Llama family in favor of proprietary approaches.

What distinguishes Inkling is not merely its technical specs but Thinking Machines' business model inversion. While OpenAI, Anthropic, and other incumbents monetize through metered API access, Thinking Machines is shifting revenue to Tinker, a platform that enables enterprises to fine-tune and run models on their own infrastructure. This approach converts the AI model itself into a commodity—developers can customize freely without per-token costs—and captures value downstream through tooling and customization. The Bridgewater collaboration exemplifies this: a financial-domain fine-tuned version achieved state-of-the-art performance on specialized benchmarks at a fraction of proprietary model costs, demonstrating that customization economics can offset raw capability gaps.

FAQ

What are the key specifications of Inkling?: Inkling is a mixture-of-experts model with 975 billion parameters, trained on about 45 trillion tokens of text, image, audio and video. For the average prompt, it draws on only about 41 billion parameters to process tasks faster and keep costs low. It can reason natively across all four input types but outputs text only, including code, styled artifacts and structured data.
How does Thinking Machines plan to make money?: Rather than charging customers for model access via API, Thinking Machines plans to generate revenue through Tinker, a paid service that makes it simple for developers to fine-tune open-weights models for specific tasks. The model itself is available to download with full open weights.
How does Inkling compare to proprietary models in cost and performance?: In early test results with Nvidia's Nemotron 3 Ultra model, Inkling achieved comparable coding performance despite using two-thirds fewer tokens. In a collaboration with Bridgewater Associates, a fine-tuned version scored 84.7% on leading financial reasoning benchmarks, outperforming the most advanced proprietary alternatives at less than 10% of the cost.

SiliconANGLE AIRead Original Article

Send on LINE

Get the latest Large Language Models news every morning

For example, today's edition would include:

Visko raises $10M, launches live AI video model OrbisSiliconANGLE AI · 2h ago
Runway unveils Solaris, an AI that generates app interfaces in real timeTHE DECODER · 2h ago
Google AI Search flags Facebook users as dangerTHE DECODER · 2h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Subscribe free →

Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →