Thinking Machines, the artificial intelligence startup founded by former OpenAI chief technology officer Mira Murati, has entered the open-weight AI race with a massive new model called Inkling. The 975-billion-parameter model is being positioned as a Western alternative to widely used open models from Chinese labs, according to Reuters.
What is Inkling?
Inkling is an open-weight large language model, meaning developers can download the model's trained parameters and run it on their own infrastructure. This approach contrasts with closed models like OpenAI's GPT-4 or Google's Gemini, which are only accessible through paid application programming interfaces (APIs).
With open-weight models, companies can fine-tune the model on their own data, adapt it to specific regulatory or privacy requirements, and potentially reduce ongoing costs compared to paying per API request. The trade-off is that running a 975-billion-parameter model requires significant computing power and technical expertise.
The model's size puts it in the same league as some of the largest open models available, though it still trails behind the biggest proprietary models from companies like OpenAI and Google.
Why a Western alternative matters
The open-weight AI space has been dominated by models from Chinese labs such as DeepSeek, Alibaba's Qwen, and Baidu's Ernie. These models have gained popularity among developers globally because they are free to use and modify, and often perform competitively with Western closed models on benchmarks.
However, some Western companies and governments have grown cautious about relying on models developed in China, citing concerns about data security, intellectual property, and alignment with local regulations. Thinking Machines aims to fill that gap with Inkling, offering a model that is open but built in the West.
This dynamic echoes broader trends in the tech industry, where geopolitical tensions are reshaping supply chains and technology partnerships. For example, Apple recently cleared a regulatory hurdle in China by tapping Alibaba and Baidu AI models for its Apple Intelligence features, highlighting how companies must navigate different regional ecosystems.
What it means for investors
For everyday investors, the launch of Inkling signals that the AI infrastructure race is far from over. The open-weight model market is becoming a key battleground, with implications for companies that provide cloud computing, chips, and data center services.
If open-weight models like Inkling gain traction, they could reduce the pricing power of API-based AI providers, potentially squeezing margins for companies like OpenAI and Anthropic. On the other hand, they could boost demand for hardware from companies like Nvidia and AMD, as running large models locally requires powerful graphics processing units (GPUs).
Investors should also watch how this affects the broader AI ecosystem. Open-weight models lower the barrier to entry for startups and smaller companies to build AI-powered products, which could accelerate innovation but also increase competition in the application layer.
Meanwhile, the push for Western alternatives to Chinese AI models could create opportunities for companies that provide AI infrastructure and services in North America and Europe. This is part of a larger trend where technology sovereignty is becoming a factor in corporate and government procurement decisions.
For now, Thinking Machines is a private company, so there is no stock to buy directly. But its success or failure could influence the strategies of publicly traded AI players and their suppliers. Investors in AI-related stocks should keep an eye on how open-weight models evolve and whether they start to capture meaningful market share from closed models.
What to watch next
The key question is whether Inkling can attract a community of developers and enterprise customers. Open-weight models live or die by their ecosystem: how easy they are to fine-tune, how well they perform on real-world tasks, and how much support they get from third-party tools and platforms.
Thinking Machines will also need to address the compute cost issue. Running a 975-billion-parameter model is not cheap, and most developers will need access to cloud GPU clusters or specialized hardware. The company may need to partner with cloud providers or offer managed services to make Inkling accessible to a broader audience.
Finally, regulatory developments could shape the open-weight landscape. Some governments are considering rules that would require AI models to undergo safety testing before release, which could affect how open-weight models are distributed and used. Investors should monitor these policy debates as they could have significant implications for the entire AI sector.
