OpenAI has expanded its model lineup with the release of GPT-5.4, available in three tiers: Standard, Thinking, and Pro. All three variants share the same underlying architecture and a 1.05 million token context window — the largest OpenAI has ever offered commercially.
Three variants explained
Standard is optimized for speed and cost, suitable for most everyday tasks. Thinking adds an extended reasoning layer for complex problems — similar to the o-series models but integrated into a single model. Pro combines maximum capability with the largest context window for the most demanding use cases.
Dynamic tool retrieval
A notable architectural change in GPT-5.4 is dynamic tool spec retrieval. Rather than loading every available tool definition into the prompt at the start of each conversation, the model fetches only the tool specifications it needs as tasks arise. This reduces prompt bloat and improves response quality in complex agentic workflows.
Availability
GPT-5.4 Standard is rolling out to ChatGPT Plus subscribers immediately. Thinking and Pro variants are available via the API and to ChatGPT Team and Enterprise customers.