Why Anthropic’s Pricing Is Triggering a Search for Alternatives
The shift away from Anthropic is a cost-driven migration in which startups and enterprises replace premium frontier AI models with cheaper open-source and alternative systems to cut inference spend while keeping acceptable or better performance. This rebalancing is reshaping how companies plan AI infrastructure, negotiate vendor contracts, and choose models for production workloads. Cost has become a breaking point. Microsoft AI CEO Mustafa Suleyman has called Anthropic “extremely expensive” and said that “many people are urgently looking for alternatives,” adding that Microsoft’s goal is to “reduce and ultimately eliminate” what it pays Anthropic. His comments underline a wider trend: as usage grows from experiments into core products, AI model invoices swell, forcing teams to question whether high-end models like Claude are worth the ongoing premium. That pressure opens the door for newer players and open source AI models that can deliver similar capabilities at lower cost.
DeepSeek vs Claude: How One Startup Claims Millions in Savings
Lindy, an AI agent platform, has become a flagship example of startups switching away from Anthropic. Founder and CEO Flo Crivello said Lindy “switched 100% of Lindy traffic to DeepSeek V4, churning from Anthropic models” and that the move “saves us millions of $ and we’re… seeing an increase in performance on many core use cases.” For a product that runs models continuously, inference had been Lindy’s largest cost, larger than payroll. DeepSeek V4-Pro is priced at USD 3.48 (approx. RM16.01) per million output tokens, and on the Artificial Analysis Intelligence Index benchmark it costs USD 1,071 (approx. RM4,930) to run, compared with USD 4,811 (approx. RM22,130) for Claude Opus 4.7. That more-than-four-times gap in AI model cost comparison scales quickly when calls reach billions per month, turning technical choices into major financial decisions.
Open Source AI Models Gain Ground with Smaller Companies
While Anthropic remains a favorite among large enterprises that can absorb higher fees for Claude and related services, smaller companies are testing open source AI models and emerging closed alternatives. Lindy evaluated several Chinese models such as GLM-5.1 and Kimi K2.5 before settling on DeepSeek V4, highlighting that this was the result of lengthy benchmarking rather than a quick reaction to hype. DeepSeek released two variants, V4-Pro and V4-Flash, and V4-Pro’s strong performance on agentic benchmarks made it attractive for AI “employee” scenarios. Startups are also building their own infrastructure stacks, as Lindy did with hosting provider Atlas Cloud, to host these models cheaply. At the same time, many still keep Anthropic in the toolbox for internal use or rare edge cases, taking advantage of subsidized plans while pushing most production traffic to lower-cost Anthropic pricing alternatives.
Enterprises, Tradeoffs, and the New Cost-Performance Frontier
The market split is becoming clear: large enterprises pay for frontier models like Claude to support complex, high-stakes workloads, while startups and mid-size firms chase better cost-performance ratios. Suleyman has said Microsoft wants to be “one of the top four labs in the world,” competing with Google DeepMind, OpenAI, and Anthropic rather than reselling them, partly to avoid high external model bills. DeepSeek itself admits it trails top frontier labs by months, not years, suggesting a narrowing gap where cost-performance tradeoff is now the central factor in model choice. For many production agentic use cases, the slight quality lag is offset by massive savings. As more companies run systematic AI model cost comparison exercises, a hybrid future seems likely, where Claude and similar models handle the hardest problems and cheaper or open models carry the bulk of everyday traffic.
