The artificial intelligence industry is experiencing a fundamental shift: instead of competing over who builds the largest model, companies are now racing to build systems that intelligently choose which model to use for each specific task. This pivot from raw model size to strategic orchestration is reshaping how businesses deploy AI and could dramatically reduce the pricing power of premium AI providers.

What's Driving the Move Away From Bigger Models?

For the past two years, the AI competition seemed straightforward: larger models, better benchmark scores, and bragging rights went to whoever launched the newest breakthrough. But as companies move from experimenting with AI to actually using it in real products and workflows, the calculus has changed entirely. The real product is no longer the model itself, but the system that decides which model to use, when to use it, and what data or tools it needs.

"The model alone is no longer the product. It is the harness, the orchestration system that puts the model inside a very capable harness and pairs the model with a lot of tools," said Aravind Srinivas, CEO of Perplexity.

Aravind Srinivas, CEO at Perplexity

This means AI products are becoming intelligent routers. A customer service chatbot might not need the most expensive model for routine questions. A complex coding problem might. A routine internal workflow could run on a cheaper open-source model, while harder tasks get escalated to a more powerful one. The goal is matching the right tool to the right job at the right cost.

How Are Open-Source Models Changing the Competition?

Open-weight models, which can be downloaded and run by companies on their own hardware, are becoming increasingly capable and significantly cheaper to operate than proprietary models from major AI labs. This shift is creating serious economic pressure on companies like OpenAI and Anthropic, which have built their business models around selling access to cutting-edge technology.

The scale of this transition could be dramatic. According to venture capital investor Peter Fenton at Benchmark, over 90% of AI tokens, the units of data that models process and generate, could come from open-weight models within the next 18 to 24 months, possibly even by the end of 2026.

"A maybe contrarian view that is becoming consensus is our belief that 90-plus percent of the tokens created will come out of open-weight models over the next 18 to 24 months, possibly even by the end of the year," said Peter Fenton, general partner at Benchmark.

Peter Fenton, General Partner at Benchmark

Many of the most competitive open-weight models are coming from Chinese AI labs, including Z.ai and DeepSeek. Perplexity recently previewed a new system built around GLM 5.2, an open model from Z.ai, designed to let cheaper models handle routine work while calling in stronger models only when necessary.

How to Implement Smart Model Routing in Your Organization

• Start With Smaller Models: Begin by deploying smaller open-source models for routine tasks running close to your own data, then expand to larger models as your team becomes more comfortable with the approach.
• Use Model Management Tools: Adopt platforms like Ollama, which makes it easier for developers and enterprises to download, run, and manage open-source models without vendor lock-in.
• Match Model Capability to Task Complexity: Reserve expensive proprietary models for genuinely difficult problems that require cutting-edge performance, while routing simpler tasks to cheaper alternatives.

Ollama, a company that simplifies downloading and running open models, has been adopted by more than 85% of the Fortune 500, including companies in heavily regulated industries like aviation, insurance, and healthcare. This widespread adoption suggests that businesses across sectors are already moving toward this hybrid approach.

"One thing is where the model's from and where it was created and trained. But the more important thing to these businesses we speak to is where it runs and how it runs," said Jeff Morgan, CEO of Ollama.

Jeff Morgan, CEO at Ollama

What Does This Mean for Data Centers and Cloud Computing?

The shift toward open models and local inference could reshape the massive data center buildout currently underway across the tech industry. The current AI boom has assumed that demand will continue flowing to large cloud data centers filled with expensive specialized chips. However, some AI work may eventually run locally on devices owned by consumers or businesses, reducing the need for constant cloud connectivity.

This doesn't mean data centers will disappear. Instead, the future likely involves a hybrid system where routine tasks run locally on devices, while the most computationally intensive work gets sent to powerful models in the cloud. This approach could reduce latency, improve privacy, and lower costs for many applications.

The economic implications are significant. As open models become more capable and companies become more selective about which models they use, the inference margins generated by frontier model companies like OpenAI and Anthropic will face increasing pressure. When businesses can run capable models without paying the markup that premium providers charge, the pricing power of those companies diminishes.

For investors and businesses alike, the key question is whether the biggest AI labs can maintain their pricing power as open models improve and companies become more sophisticated about cost optimization. The answer appears to be shifting in favor of a more distributed, cost-conscious AI ecosystem.