GPT-5.4 Pro — Specs, Pricing & Limits
Specs
| Status | Priced |
|---|---|
| Context window | 1,050,000 tokens |
| Max output | 128,000 tokens |
| Knowledge cutoff | Aug 31, 2025 |
| Reasoning effort levels | medium (default), high, xhigh |
| Long-context rule | >272,000 input tokens: 2x input, 1.5x output (full request) |
| Cache writes | No additional charge — a written token bills as ordinary input |
Pricing per 1M tokens, every published service tier
| Standard — short context | in $30.00 · cached — · writes — · out $180.00 |
|---|---|
| Standard — long context | in $60.00 · cached — · writes — · out $270.00 |
| Batch — short context | in $15.00 · cached — · writes — · out $90.00 |
| Batch — long context | in $30.00 · cached — · writes — · out $135.00 |
| Flex — short context | in $15.00 · cached — · writes — · out $90.00 |
| Flex — long context | in $30.00 · cached — · writes — · out $135.00 |
| Fast mode | Not offered for this model |
Rate limits by usage tier
| Free | Not supported |
|---|---|
| Build | 500 RPM · 100,000 TPM · 1,350,000 batch queue |
| Launch | 1,000 RPM · 400,000 TPM · 200,000,000 batch queue |
| Grow | 1,500 RPM · 4,000,000 TPM · 15,000,000,000 batch queue |
Long-context rate limits (requests over 272,000 input tokens, in place of the table above)
| Free | Not supported |
|---|---|
| Build | 50 RPM · 100,000 TPM · 20,000,000 batch queue |
| Launch | 200 RPM · 1,000,000 TPM · 100,000,000 batch queue |
| Grow | 800 RPM · 2,000,000 TPM · 1,000,000,000 batch queue |
GPT-5.4 Pro shares the largest Standard sticker price on this site's roster with GPT-5.5 Pro, and the most important thing to know about it still isn't a price: its entry-tier token ceiling doesn't scale with its context window the way the price suggests it should.
That mismatch got sharper, not softer, through the autumn. OpenAI first cut this model's request rates, then on October 6, 2026 re-published its table for the Build, Launch and Grow ladder, where its Build token ceiling is the smallest of any model this site prices. Its page also publishes a separate long-context table for prompts over 272,000 input tokens, with far fewer requests per minute — and a maximum-length prompt is exactly such a request. The window itself matches the GPT-6 Sol and Astra rows, so one maximum-length prompt takes many minutes of Build-tier token budget to clear. The rate-limit tier planner computes the exact figure from the table above — plan the account tier alongside the window, not after the first throttled request in production.
It publishes the long-context repricing rule for Standard, Batch and Flex, and offers no Fast mode and no cached-input price. Its reasoning effort runs medium, high and xhigh, with medium as the default. Like GPT-5.5 Pro, its requests can take minutes, and OpenAI suggests background mode.
For Pro-style extra model work on a newer model, reasoning.mode: "pro" on GPT-5.6 and GPT-6 bills at the selected model's own standard rates.
The tables above are computed directly from this site's facts module: pricing across every published service tier, the long-context bands, and both rate-limit tables — read the rate-limit rows before assuming price is the only constraint.
Verified 2026-10-28 against https://developers.openai.com/api/docs/models/gpt-5.4-pro.
Checked: https://developers.openai.com/api/docs/models/gpt-5.4-pro · https://developers.openai.com/api/docs/pricing