OpenAI and Anthropic slash LLM prices by up to 80% in months—why?
anthropic openai
| Source: HN | Original article
OpenAI and Anthropic have reduced their language‑model API prices by up to 80% within months, prompting industry speculation about the reasons.
OpenAI and Anthropic have slashed the fees for their language‑model APIs by roughly half to eight‑tenths within a few months of 2026, marking one of the steepest price adjustments the sector has seen. OpenAI’s GPT‑5.6 “Luna” model fell 80 % in July, while its Opus 5.5 offering trimmed cache‑read costs enough to lower total pricing by 40‑60 % versus Opus 5. Anthropic’s latest Fable 5.1 is now billed at $10 / $50 / $0.25, a 75 % reduction in cache‑read cost compared with the June‑released Fable 5, which was priced $10 / $50 / $1.
The cuts are being framed as a response to efficiency gains and mounting competitive pressure. Both firms point to deeper optimisation of inference pipelines—particularly cache handling—as the primary driver of lower per‑token costs. Analysts also note that the moves intensify a nascent price war, echoing earlier market shifts such as Replit’s recent migration of most AI spend from Anthropic to OpenAI after Anthropic discontinued a discount and OpenAI offered a 30 %‑plus reduction (see our coverage on 25 Oct 2026). By making their services markedly cheaper, the two companies hope to lock in developers, startups and research teams that might otherwise look to emerging rivals or to open‑source alternatives.
The ramifications will unfold across the AI ecosystem. Developers can now run larger or more frequent queries without inflating budgets, potentially accelerating product launches and experimentation. At the same time, the steep discounts raise questions about the sustainability of profit margins for frontier‑model providers, especially as they continue to train ever larger systems under tighter security constraints—a theme we explored in our 25 Oct 2026 reports on OpenAI’s offline training regime.
Going forward, observers will watch whether the price trajectory continues, how rival firms such as Microsoft, Google or emerging open‑source platforms react, and whether the cost reductions translate into measurable gains in market share or usage volume. Further announcements on model performance, new pricing tiers or additional efficiency breakthroughs will be key indicators of how deep the current price war will run.
Sources
Back to AIPULSEN