AI firms rush to cheaper open‑weight models as pricing competition heats up

U.S. companies are increasingly pairing high‑performance AI models such as OpenAI's GPT‑5 and Anthropic's Opus with lower‑cost, open‑weight alternatives to reduce operational expenses. A test run showed that using GPT‑5.5 alone cost over $10,000 for 50,000 tokens, while a mix of a cheaper composer model and Anthropic’s Opus 4.8 totaled just $1,339. Companies like Filon have secured millions of dollars in free token credits and unlimited‑use offers from AI vendors, prompting a shift away from long‑term loyalty to any single provider.

At the same time, leading U.S. tech firms—including Nvidia, Microsoft, Meta, OpenAI and Google—have signed an open letter opposing new regulations on open‑weight AI models. They warn that attempts to curb China’s AI progress could inadvertently shrink the U.S. open‑AI ecosystem, arguing that publicly available model weights enhance safety, innovation and cost‑effectiveness across sectors such as manufacturing, healthcare and education.