Why Open Source AI Rise Isn't Hurting Anthropic Yet: Dual-Track AI Economy Forms
📌 Key Takeaways
- • Decagon CEO Jesse Zhang proposes new theory: frontier and open source models aren't competitors but two phases of the same life cycle
- • DeepSeek has surged to lead in token usage on Vercel platform, processing over one-third of tokens
- • But Anthropic still accounts for over half of AI spend on Vercel platform
- • OpenRouter data: DeepSeek V4 Flash processes 5.3 trillion tokens weekly, Opus 4.8 handles 2T+, but Opus unit price is 23x V4 Flash
- • Frontier models dominate “discovery phase,” open source models dominate “production phase,” forming stable dual-track economy
In July 2026, an interesting dual-track economy is forming in the AI market. On one hand, open source models like DeepSeek are growing wildly in usage; on the other, frontier models like Anthropic's Opus 4.8 still firmly hold the lion's share of spending. Behind this seemingly contradictory phenomenon lies the deep logic of AI industry development.
I. The “Dual-Track” Theory
On Monday, Decagon CEO Jesse Zhang published a provocative new article titled “Everyone is wrong about open source AI in the enterprise.” The post grapples with one of the most interesting contradictions of today's AI economy: more mature AI deployments are switching to lighter models, but overall spend on expensive state-of-the-art models has barely budged.
Zhang's theory offers a fresh perspective on the relationship between frontier and open source models. In his telling, they aren't competitors, and open source models' success isn't coming at the expense of frontier labs. Instead, they're two phases of the same life cycle—expensive frontier models are used to prove out use cases that can be passed along to cheaper open source alternatives as they mature.
💡 Core Logic of Dual-Track Economy
- • Discovery Phase: Frontier models dominate, used for exploring new use cases, validating feasibility
- • Production Phase: Open source models dominate, used for scaled deployment, cost reduction
- • Dynamic Balance: New use cases continuously emerge, frontier models maintain high demand
- • Price Difference: Frontier model unit price is 10-23x that of open source models
- • Spending Structure: Frontier models account for the majority of spending, despite potentially lower usage
II. Data-Supported Dual Track
Vercel's AI gateway dashboard shows that in just the past week, DeepSeek has surged into the lead for token volumes, now processing just over a third of the tokens passing through the company's infrastructure. Z.ai—the lab behind the popular GLM-5.2 model—jumped into a respectable fourth place over the same period.
But if you scroll down to overall token spend, you'll see Anthropic still accounts for more than half of the overall AI spend on the platform. Given that much of the recent change comes from Anthropic's own rising prices, the share has dropped slightly over the past month, but not significantly.
OpenRouter tells a similar story, capturing a much larger (but slightly less enterprise-y) segment of the market. DeepSeek V4 Flash is the main winner on overall usage, processing 5.3 trillion tokens weekly. The most popular frontier model, Opus 4.8, handles just over 2 trillion. OpenRouter doesn't rank models by total spend, but it registers the average token cost for Opus 4.8 as roughly 23x higher than V4 Flash ($1.37 per million tokens, compared to just 6 cents), which would mean Opus was still probably capturing the lion's share of spending.
III. Why Frontier Models Maintain Position
Those figures don't fully prove Zhang's point about the AI life cycles, but they do show frontier labs like Anthropic aren't suffering too much from the rise of open source. One explanation is that the market of AI-addressable tasks is growing so fast that the top models are able to maintain their position just by dominating early-stage deployments. As Zhang puts it, “The frontier labs will keep owning discovery. Open source will increasingly own production.”
Another explanation might be that, even as clients move to open source, many use cases are so difficult that they can't be entirely replaced with cheaper alternatives. Frontier models still have advantages in handling complex reasoning, multi-step tasks, and edge cases—areas where open source models currently struggle.
🔧 Related Tools
AI Content Generator
Generate high-quality content using latest AI models, supporting multiple scenarios
AI Code Assistant
Intelligent code generation and debugging tool to boost development efficiency
Text Summarizer
Quickly extract core content from long documents, saving reading time
IV. New Players Enter
Those figures don't even capture the newest arrival—Nvidia's Nemotron, which is “poised to leap to the front of the pack” by virtue of Nvidia's strong connections and the model's own extreme adaptability. This shows the AI market's competitive landscape is still rapidly changing, with new variables constantly entering.
Nvidia, as the dominant player in the GPU market, has natural advantages entering the AI model market. It can not only optimize models to run better on its own hardware but also leverage its massive customer base for rapid promotion. Nemotron's arrival may further complicate the dual-track economy landscape.
V. Implications for Enterprises
For enterprises, understanding this dual-track economy is crucial. First, don't simply assume “open source is better” or “frontier is better”—they serve different purposes. Second, establish flexible model strategies, choosing appropriate models based on use case maturity.
Third, focus on total cost of ownership (TCO), not just token prices. Frontier models, while expensive per token, have higher success rates on complex tasks, which may mean fewer retries and lower overall costs. Finally, stay attuned to new models, as the market landscape changes quickly—today's best choice may not be optimal tomorrow.
Frequently Asked Questions (FAQ)
Q1: What is the “dual-track” AI economy?
The dual-track AI economy refers to a landscape where frontier and open source models play different roles in the AI market. Frontier models (like Anthropic Opus 4.8, OpenAI GPT-5) are used for exploring new use cases, handling complex tasks, and account for the majority of spending; open source models (like DeepSeek V4, Z.ai GLM-5.2) are used for scaled deployment, handling mature use cases, and account for the majority of usage. The two form a complementary rather than competitive relationship.
Q2: Why do open source models have high usage but low spending?
This is because open source model token prices are far lower than frontier models. Take OpenRouter data as an example: Opus 4.8's average token cost is 23x that of DeepSeek V4 Flash ($1.37/million tokens vs $0.06/million tokens). Therefore, even though DeepSeek processes 2.5x the tokens of Opus, Opus's spending remains higher.
Q3: How should enterprises choose models?
We recommend a layered strategy: 1) For new use cases, complex reasoning, and critical tasks, use frontier models to ensure quality; 2) For mature use cases, simple tasks, and large-scale deployment, use open source models to reduce costs; 3) Establish model evaluation mechanisms, regularly review use case maturity, and migrate from frontier to open source models when appropriate; 4) Maintain multi-model architecture to avoid vendor lock-in.
Q4: How long will this dual-track system last?
In the short term (1-2 years), the dual-track system is likely to persist, as the market of AI-addressable tasks is still rapidly expanding and new use cases continuously emerge. Long-term, if open source model capabilities continue improving rapidly and can handle more complex tasks, the dual-track boundaries may blur. But given frontier labs' continuous innovation, complete replacement by open source is unlikely. More likely is evolution to a multi-track system—models at different price points and capability levels serving different scenarios.
Conclusion
The rise of open source AI hasn't hit frontier labs as hard as many predicted. Instead, the AI market is forming a more complex and stable dual-track economy. Frontier models dominate “discovery,” open source models dominate “production,” forming a complementary rather than substitutive relationship.
This landscape means enterprises need more refined model strategies—not simply choosing “cheapest” or “strongest,” but selecting appropriate models based on use case maturity and complexity. For frontier labs, the key is continuous innovation, maintaining leadership in the “discovery phase,” while preserving pricing power through service differentiation and ecosystem building.
For the entire AI industry, the formation of a dual-track economy is a positive signal—it means AI is moving from “technology demonstration” to “scaled application,” from “single model” to “diversified ecosystem.” This maturity will drive broader AI technology adoption, ultimately benefiting society as a whole.