Why AI Model Costs Are Suddenly Under the Microscope
The shift from premium AI providers to cheaper or open source AI models is a growing trend in which companies replace high-priced, closed systems with more affordable alternatives that still meet performance needs, aiming to reduce inference costs that have become one of their largest ongoing expenses. This change is driven by how often AI models are called in production products, where every million tokens processed adds up. For many AI-focused startups, inference bills rival or exceed payroll, turning AI model cost comparison into a core business task rather than a side concern. As a result, teams are reevaluating their dependence on a few frontier labs and building room in their architectures to swap models when pricing or terms no longer make sense.
Lindy’s Big Switch: From Anthropic to DeepSeek
Nothing illustrates the pressure on Anthropic pricing alternatives better than Lindy’s move. Flo Crivello, CEO of AI agent platform Lindy, announced that the company has “switched 100% of Lindy traffic to DeepSeek v4, churning from Anthropic models,” saying the change “saves us millions of $ and we’re seeing an increase in performance on many core use cases.” For Lindy, inference had become the number one cost line, larger than payroll, so a 2–5x reduction was described as “transformative” for the business. DeepSeek V4-Pro’s listed price of USD 3.48 (approx. RM16) per million output tokens, and its benchmark cost of USD 1,071 (approx. RM4,930) on the Artificial Analysis Intelligence Index versus USD 4,811 (approx. RM22,160) for Claude Opus 4.7, shows how DeepSeek vs Anthropic comparisons now tilt toward lower-cost models for large-scale workloads.
When Even Microsoft Says Anthropic Is Too Expensive
The cost squeeze is not limited to startups. In an interview with Bloomberg, Microsoft AI CEO Mustafa Suleyman described Anthropic as “extremely expensive” and said “many people are urgently looking for alternatives.” He added that Microsoft itself “pay a lot of money to Anthropic,” and that the company’s goal is to “reduce and ultimately eliminate that cost.” That statement matters because Microsoft is one of Anthropic’s major enterprise customers and has deep visibility into enterprise AI savings opportunities. Suleyman framed this not merely as cost cutting but as a competitive move: Microsoft aims to be one of the “top four labs in the world,” building its own frontier models instead of depending on partners like Anthropic or OpenAI. Cost is becoming a strategic driver of which labs enterprises are willing to depend on.
Open Source AI Models as Enterprise-Grade Replacements
The Lindy case shows that open source AI models and open-weights alternatives are no longer experimental toys. DeepSeek V4-Pro scores 1554 on the GDPval-AA agentic benchmark, leading open-weights models on real-world tasks at launch, which is the kind of workload many enterprise AI deployments care about. Lindy reports that performance on core tasks went up after moving away from Anthropic, undermining the idea that cheaper always means worse. At the same time, the team still keeps Claude available for internal use and rare edge cases, taking advantage of an “absurd max plan subsidy.” This hybrid pattern is likely to spread: open source and lower-cost models handle the bulk of traffic, while premium proprietary models sit at the edges for specialised or high-stakes uses, reshaping AI model cost comparison in practice.
How Cost Comparisons Are Reshaping AI Vendor Choices
For both startups and large enterprises, AI model cost comparison is becoming a routine part of vendor selection, not an afterthought. When one benchmark can be run for USD 1,071 (approx. RM4,930) on DeepSeek V4-Pro and USD 4,811 (approx. RM22,160) on a premium model like Claude Opus 4.7, procurement teams notice. The ratio compounds when traffic runs into billions of tokens a month. According to Lindy’s CEO, the company’s migration took “100x more work than we thought,” including custom infrastructure and tooling, but the ongoing savings and performance gains justified the effort. With Microsoft’s AI chief openly targeting a reduction in what his company pays Anthropic, the signal to the market is clear: cost-conscious organizations are expected to pressure providers, consider open alternatives, and treat AI model choice as a financial decision as much as a technical one.
