OpenAI Enterprise usage-based billing

OpenAI has published a token-based rate card for Enterprise customers and is transitioning them from a fixed-credits model to usage-based token pricing. The rate card covers GPT-5.6 Sol, Terra, and Luna tiers, GPT-5.5 variants, GPT-5.4, GPT-5.3-Codex, deep-research models, and image generation, with ChatGPT for PowerPoint pricing shifting to tokens after August 6, 2026. Codex moved to usage-based billing on April 2, 2026, where heavy agentic coding workflows trigger separate billing through workspace credits beyond the baseline in standard Enterprise seats; variable billing is described as difficult to forecast without pilot data. Alongside this token-pricing transition, OpenAI is also exploring a further step: task-based pricing where enterprise clients pay only for successfully completed tasks rather than for tokens consumed, a model described as potentially far more lucrative though logistically complex. Rohan Paul argued in parallel that token pricing has become an incomparable unit across models, proposing cost per successful task as the better metric, and cited GPT-6 Astra as an example cheaper per task than Gemini 3.8 Flash despite Flash carrying a 13x lower nominal token price.