OpenAI introduced GPT-5.4 across ChatGPT, Codex, and API
OpenAI launched GPT-5.4 on March 5, 2026, with rollout to paid ChatGPT users, all Codex users, and immediate API availability. OpenAI also exposed GPT-5.4 Pro in the API, pricing GPT-5.4 at $2.50 input / $0.25 cached input / $15 output per 1M tokens and GPT-5.4 Pro at $30 input / $180 output per 1M tokens.
Impact Assessment
- New OpenAI launch: GPT-5.4
- Companion API variant: GPT-5.4 Pro
- Rollout started in ChatGPT for paid users and in Codex for all users
- API pricing: GPT-5.4 at $2.50 input / $0.25 cached input / $15 output per 1M tokens
- API pricing: GPT-5.4 Pro at $30 input / $180 output per 1M tokens
- Context window: 1.05M tokens, max output: 128K tokens
- OpenAI said GPT-5.2 Thinking remains available in ChatGPT until June 5, 2026
- Benchmark GPT-5.4 against GPT-5.2 and your current high-intelligence routing targets on long-context and professional workflows.
- Update cost guardrails because GPT-5.4 Pro is materially more expensive and long prompts above 272K tokens carry higher pricing multipliers.
- If your team relies on GPT-5.2 Thinking in ChatGPT, plan internal migration before June 5, 2026.
Changes
| Field | Before | After |
|---|---|---|
| models | — | gpt-5.4, gpt-5.4-pro |
| chatgpt_rollout | — | Plus, Pro, Team |
| codex_rollout | — | all_users |
| api_pricing | — | { "gpt-5.4": { "input_per_1m": 2.5, "cached_input_per_1m": 0.25, "output_per_1m": 15 }, "gpt-5.4-pro": { "input_per_1m": 30, "output_per_1m": 180 } } |
| context_window | — | 1050000 |
| max_output | — | 128000 |
Recommended Actions
TEST
Benchmark GPT-5.4 on your long-context and professional-workload eval set before changing default routing.
MONITOR
Add separate budget caps for GPT-5.4 Pro because its input and output token pricing is materially above standard GPT-5.4.
MIGRATE
If your team still uses GPT-5.2 Thinking in ChatGPT workflows, plan a switch before the June 5, 2026 availability deadline.
Sources
OpenAI says GPT-5.4 starts rolling out to paid ChatGPT users today and to all users in Codex.
The GPT-5.4 model page lists a 1,050,000 token context window, 128,000 max output tokens, and $2.50 input / $0.25 cached input / $15 output pricing.
The GPT-5.4 Pro model page lists $30 input / $180 output pricing and the same 1,050,000 token context window.
OpenAI API pricing lists GPT-5.4 and GPT-5.4 Pro token prices and notes higher pricing for prompts above 272K tokens.