The New Economics of AI Models
The current wave of AI infrastructure buying is pushing companies to compare model accuracy against inference cost, making AI model cost comparison a central strategic decision for startups that run models at scale in production environments. Instead of choosing a single premium provider and absorbing the bill, founders are now building stacks that mix closed-source and open source AI models, tuned for specific workloads and price points. That shift is changing how leaders think about AI cost savings: inference spend is no longer a fixed tax, but a variable they can control through model choice, deployment architecture, and hosting providers. In this landscape, open-weight models like DeepSeek V4 are emerging as a credible DeepSeek alternative to premium services, especially when performance is close enough that cost becomes the deciding factor.
Lindy’s DeepSeek Switch: Millions Saved, Performance Up
AI agent platform Lindy is now the clearest example of this trend. Founder Flo Crivello says the company has switched all customer traffic from Anthropic models to DeepSeek V4, describing the move as “transformative for the business.” For Lindy, inference was the single biggest line item, larger than payroll, so a cheaper model mattered more than any marginal quality gain. Crivello reports that the shift “saves us millions of $ and we’re actually seeing an increase in performance on many core use cases,” turning price into upside instead of compromise. On benchmarks, DeepSeek V4-Pro is priced at USD 3.48 (approx. RM16.01) per million output tokens and costs USD 1,071 (approx. RM4,932.60) to run the full Artificial Analysis Intelligence Index, compared with USD 4,811 (approx. RM22,141.40) for Claude Opus 4.7.
Why Anthropic Pricing Is Under Pressure
Lindy’s migration did not happen in isolation; it landed in a market where Anthropic pricing is already under scrutiny. In a Bloomberg interview, Microsoft AI CEO Mustafa Suleyman said, “Anthropic is extremely expensive, and I think many people are urgently looking for alternatives,” adding that Microsoft’s goal is “to reduce and ultimately eliminate” what it pays them. That is unusually blunt language from a major enterprise customer. Microsoft’s ambition is to build its own frontier models rather than rely on partners, underscoring how price and strategic control now go together. When both startups and large platforms say closed models strain their budgets, it signals that premium AI providers may need to revisit their economics, or risk watching workloads drift toward open-source AI models and more flexible pricing.
Open-Source AI Models Become Viable Defaults
DeepSeek V4’s appeal is not price alone; it is also about where performance now stands. V4-Pro scores 1554 on the GDPval-AA benchmark focused on real-world agent tasks, making it the leading open-weights model on that test at launch. DeepSeek itself says it trails the US frontier by about 3–6 months, yet for many agentic use cases that delay no longer matters. Lindy spent months benchmarking rivals such as GLM-5.1 and Kimi K2.5 before choosing DeepSeek, and still keeps Anthropic’s Claude internally for subsidized usage and rare fallback cases. This pattern shows how a DeepSeek alternative can become the default for high-volume workloads, while premium models are reserved for edge cases. For cost-conscious teams, the default assumption is shifting: closed models must now justify their price in clear, measurable gains.
Market Implications: Pricing Power Meets Open Competition
As more companies run careful AI model cost comparison exercises, the market is tilting toward diversified stacks rather than single-vendor dependence. Microsoft sees a large untapped market where most people do not yet use AI daily, which means pricing decisions made now will shape who wins those future users. If open and open-source AI models keep closing the performance gap while staying 2–4 times cheaper on real workloads, business buyers will treat expensive models as a luxury, not a default. Enterprise relationships may keep Anthropic strong at the top of the market, but the gravity of AI cost savings is pulling startups and even big platforms toward cheaper, capable alternatives. The message to premium labs is clear: frontier performance alone is no longer enough; the bill has to make sense too.
