Starting in December, Plus, Pro, and Business plans will share a single token budget across ChatGPT and Codex. As OpenAI folds Work's agentic features into the main interface, the operational mechanics of enterprise AI consumption are undergoing a fundamental architectural restructuring. OpenAI had already announced plans to merge Chat and Work, shifting away from the flat-rate consumption model toward a resource-governed framework.
Tokens measure the information a model processes and generates, so more complex models and tasks eat through the budget faster. This transition dismantles the predictability of fixed-fee software accounting, replacing unlimited access assumptions with strict consumption boundaries. As enterprise spending models shift from predictable SaaS seat licenses to dynamic utility billing, corporate finance departments face a new challenge. Can engineering teams accurately forecast financial expenditure when advanced agentic loops consume resources at unpredictable velocities?
Users can check their remaining balance and reset times in Settings, and once they hit the limit, they can wait for a reset, upgrade, or buy extra credits. OpenAI recently doubled usage limits for higher-priced plans, presumably to prepare for the merger, which itself is likely a response to criticism of ChatGPT's growing complexity. Users currently choose between two operating modes and 11 model settings, creating an interface burden that Anthropic already addressed by combining Chat and Cowork in Claude.
Chats with Instant models won't count toward the budget, so on paper ChatGPT stays a flat-rate service, but heavy usage on cheaper plans will likely mean lower-quality results. This dual-track structure masks the true depletion rate of enterprise token allocations behind unmetered lightweight interactions, forcing organizations to re-evaluate how they budget for variable computing resources.
Check account remaining balances and reset schedules inside Settings to audit baseline token consumption patterns before deployment limits restrict active workflows.
The catch
A single shared token pool across ChatGPT and Codex replaces predictable flat rates with strict consumption caps. OpenAI streamlines an interface overloaded with 11 model settings, but shifts the raw cost of compute-heavy tasks directly onto buyers. Teams face an awkward trade-off: settle for lower-quality responses from lightweight Instant models once limits run out, or constantly buy extra credits to keep autonomous agent loops running. Packaging chat and agentic workflows into one interface effectively turns software access into an unmetered risk where every deep autonomous run chips away at a hard computational limit.