Open Source AI Growth Not Hurting Frontier Labs Like Anthropic, Yet

The rapid expansion of open-source AI models is not yet significantly impacting the market share or revenue of leading frontier AI labs like Anthropic, according to a recent theory put forth by Decagon CEO Jesse Zhang. This perspective challenges the conventional idea that open-source alternatives directly compete with and erode the market of more expensive, state-of-the-art models. Instead, Zhang suggests a symbiotic relationship where both types of models cater to different stages of AI adoption.

Zhang’s theory, detailed in his post “Everyone is wrong about open source AI in the enterprise,” posits that mature AI deployments often transition to lighter, more cost-effective open-source models once use cases are proven. However, the overall expenditure on frontier models remains robust because new, complex use cases continuously emerge, requiring the advanced capabilities of these cutting-edge systems. This creates a lifecycle where frontier models drive discovery, and open-source models increasingly handle production.

Market Data Supports Two-Tiered Economy

Although Zhang’s initial post lacked extensive data, market observations from platforms like Vercel and OpenRouter largely corroborate his theory. Vercel’s AI gateway dashboard shows DeepSeek leading in token volumes, processing over a third of tokens on its infrastructure recently. Z.ai’s GLM-5.2 also secured a strong fourth place. Yet, in terms of overall token spend on Vercel, Anthropic still commands more than half, indicating a significant premium.

OpenRouter, another major platform, reveals a similar trend. DeepSeek V4 Flash is a dominant force in usage, handling 5.3 trillion tokens weekly, while the popular frontier model, Opus 4.8, processes just over 2 trillion. However, Opus 4.8’s average token cost is approximately 23 times higher than V4 Flash ($1.37 per million tokens compared to 6 cents). This substantial price difference suggests that despite lower volume, frontier models like Opus 4.8 still capture a significant portion of spending due to their premium pricing.

Frontier Labs Own Discovery, Open Source Owns Production

This emerging two-tiered economy suggests that frontier labs will continue to drive innovation and discovery for novel AI applications. Their advanced models are crucial for initial proof-of-concept and tackling complex, unsolved problems. As these use cases mature and become more standardized, enterprises can then switch to more economical open-source alternatives for large-scale production.

The market for AI-addressable tasks is expanding at such a rapid pace that even as clients migrate to open-source solutions for established tasks, new demands keep frontier models in high demand. This dynamic allows top models to maintain their position by dominating early-stage deployments where their sophisticated capabilities are indispensable.

Implications for the AI Ecosystem in India

For the Indian AI ecosystem, this trend has significant implications. Startups and enterprises can leverage open-source models for cost-effective scaling of proven AI applications, fostering broader adoption and innovation. Simultaneously, a thriving market for frontier models means that highly complex, specialized AI challenges – relevant to sectors like healthcare, finance, or advanced manufacturing in India – will continue to attract investment and development from leading AI labs. This dual approach could accelerate AI integration across various industries, balancing cutting-edge research with practical, scalable deployments.

Key facts

Feature	Frontier Models (e.g., Anthropic Opus 4.8)	Open Source Models (e.g., DeepSeek V4 Flash)
Primary Role	Discovery, complex use cases	Production, mature use cases
Cost per Token	Significantly higher	Significantly lower
Market Share (Spend)	High (premium pricing)	Lower (high volume)

Source: TechCrunch AI – https://techcrunch.com/2026/07/07/why-the-rise-of-open-source-ai-isnt-hurting-anthropic-yet/