OpenAI's GPT-5.6 line and enterprise ChatGPT push annual recurring revenue past a full prior quarter — as Codex adoption climbs
OpenAI's annual recurring revenue has now surpassed the whole of a prior quarter's run, credited to the GPT-5.6 series (Sol, Luna, Terra), the enterprise ChatGPT tier, and rising use of the Codex coding tool. The capability story and the revenue story have converged: the models that move benchmarks are the same ones moving the P&L.
The GPT-5.6 split into Sol, Luna, and Terra is a segmentation move as much as a capability one — a tiered lineup aimed at pricing different workloads rather than one model priced for all of them. The revenue lift comes less from a single benchmark win than from matching model cost to task value across a portfolio, which is how you grow ARR without giving margin away on cheap queries.
Codex is the quieter driver. Coding assistance has turned into one of the highest-retention enterprise AI use cases because it produces artifacts a team keeps using, and OpenAI's traction there mirrors what has propelled Anthropic's Claude Code. The coding tool is now a strategic surface, not a demo.
The context is a valuation race OpenAI is, for the moment, losing to Anthropic — which makes the revenue print a prospectus exhibit as much as a product update. A fall IPO would put OpenAI's 2025 losses on the same page as a rival claiming a near-term operating profit, and the GPT-5.6 revenue is the counterweight OpenAI needs on that page.
LLM-Stats — LLM news today (August 2026) → · CNBC — Anthropic tops OpenAI as most valuable AI startup →