June 24, 2026
Luxury AI meets bargain-bin chaos
The Unbearable Cheapness of Open Weight Models
AI’s fancy brands may have a cheap-model problem, and the comments are ruthless
TLDR: The post argues that some open AI models are now so cheap they make big-name tools look overpriced by comparison. In the comments, people piled on with a brutal question: if cheaper options are already good enough for most jobs, how long can the premium AI giants keep charging luxury rates?
The big gasp in this story is simple: some newer, more open AI models are wildly cheaper than the glossy names people know, with the post claiming the gap can hit nearly 50 times on usage costs alone. That immediately sent the community into full popcorn mode. The hottest reaction was basically, “If cheaper models are already good enough, why are people still paying luxury prices?” Several commenters compared OpenAI and Anthropic to designer labels trying to sell exclusivity, not just software. In other words: less “best tool wins,” more “VIP rope outside the club.”
And the comment section did not stop there. One camp argued this is just the normal march of tech: today’s expensive miracle becomes tomorrow’s bargain bin superstar. Another camp was much harsher, saying the big AI labs may have a brutal business problem on their hands if lower-cost rivals keep improving. One user bragged that tools built on these cheap models let them do a month of work for under two bucks, which is the sort of comment that starts online stampedes. Others predicted the giant companies will have to justify their prices with extra services, safety promises, or custom support, because raw AI answers are starting to look like a commodity.
The spiciest paranoia? Fear that the premium players may lean on politics and security fears to slow open competitors. That claim is speculative, but it lit up the thread because it turns a pricing debate into a full-on power struggle. The mood was a mix of glee, skepticism, and “the emperor has no tokens” energy.
Key Points
- •The article says DeepSeek V4 is priced far below frontier models from Anthropic and OpenAI, with a token-cost gap described as nearly 50x.
- •The article discusses whether low-cost open-weight models are cheap because of operational efficiency across diverse hardware or because providers are pricing them aggressively.
- •It states that Google released Gemma 4 in April 2026, OpenAI last released open-weight GPT models in 2025, and Anthropic has not released an open-weight model to the author’s knowledge.
- •The article distinguishes open weight from fully open-source AI by emphasizing openness of the training data pipeline in addition to model weights.
- •Allen AI’s OLMo models are presented as downloadable examples of more open AI, and the article cites an NSF-Nvidia partnership to support Allen AI in developing a fully open AI system.