An engineer critiques the current AI market, arguing that high-cost frontier models from companies like Anthropic and OpenAI are being outperformed by efficient alternatives like Qwen 3.7 Max.

Key Points

• The author argues that US-based frontier models are plateauing while charging excessive premiums compared to open-source or international alternatives.
• Qwen 3.7 Max is highlighted for its native extended-thinking capabilities and reliability in long-duration, complex engineering tasks.
• Excessive corporate spending on AI, including multi-million dollar monthly token bills, is criticized as a wasteful trend driven by hype rather than value.
• The author suggests that developers are increasingly shifting toward cost-effective providers like DeepSeek, Moonshot, and MiniMax via platforms like OpenRouter.
• Benchmarking data from Artificial Analysis and real-world developer usage patterns are cited as evidence that the "intelligence gap" between models is narrowing.

Why it Matters

The rapid adoption of expensive, proprietary AI models by corporations risks significant financial instability if these investments fail to deliver tangible productivity gains. This shift toward more affordable, high-performance alternatives suggests a potential market correction that could impact inflated IPO valuations and broader investment portfolios.