Anthropic leads in AI model intelligence with Fable 5 but struggles with capacity constraints due to past underinvestment in compute infrastructure, limiting user quotas and risking losing customers to OpenAI’s more generous and cost-effective GPT 5.6. Meanwhile, OpenAI’s early infrastructure investments and user-friendly approach have secured developer preference and market advantage, though Anthropic hopes recursive self-improvement will help it regain dominance in the future.
Anthropic is currently leading the AI race against OpenAI in terms of revenue and model intelligence, boasting the smartest model on the planet, Fable 5. However, a critical mistake made two years ago—hesitation to heavily invest in compute infrastructure—continues to haunt Anthropic. This cautious approach, intended to avoid bankruptcy if AI demand didn’t meet expectations, backfired as demand surged beyond predictions. Consequently, Anthropic struggles with capacity constraints, limiting their ability to offer generous usage quotas for Fable, unlike OpenAI, which aggressively invested in compute resources early on.
OpenAI has capitalized on this by frequently resetting user quotas, allowing subscribers to maximize their token usage repeatedly, a practice that has become almost a meme within the community. In contrast, Anthropic’s quotas are less generous and rarely reset, and they are considering removing Fable from their subscription plans, making it accessible only via expensive API usage. This move risks driving users toward OpenAI’s more cost-effective and efficient GPT 5.6 model, which, while slightly less intelligent than Fable, offers better value per task and more generous quotas, making OpenAI subscriptions more attractive overall.
Despite these challenges, Anthropic is aware of the risks of removing Fable from subscriptions and has extended access multiple times, signaling their intent to restore it once capacity allows. The video highlights the stark contrast in user experience and company approach: OpenAI is praised for transparency, generosity, and a user-friendly attitude, while Anthropic is perceived as less caring toward its users. This difference influences developer preference, with many leaning toward OpenAI due to better service and reliability, despite Anthropic’s superior technology.
Looking ahead, the two models represent different stages of development. GPT 5.6 is a mature, reliable “veteran” model nearing the end of its generation, while Fable is a “rookie” with immense potential for growth and efficiency improvements. OpenAI is likely preparing for GPT-6, which may be larger and more costly to serve, potentially facing similar supply challenges as Anthropic. However, OpenAI’s early investment in infrastructure positions them well to handle future demand, whereas Anthropic must expand capacity to remain competitive.
Finally, the video discusses the concept of recursive self-improvement, where the best AI model could iteratively improve itself, potentially widening the gap between Anthropic and OpenAI. Anthropic might be betting on this phenomenon to maintain long-term dominance despite current setbacks. Meanwhile, developer sentiment and user experience play crucial roles in market dynamics, with OpenAI currently winning the emotional and practical battle. The recommendation for users is to choose OpenAI subscriptions for better value and reliability while continuing to watch how Anthropic evolves and addresses its capacity issues.