GPT-5.4 Pro
GPT-5.4 Pro uses more compute to think harder and provide consistently better answers.
text + image + file → text · made by OpenAI
Sold by 19 ways
| Seller | Lane | Rate |
|---|---|---|
| Microsoft Azure AI cloud | flex | $15per Mtok in$90per Mtok outraw.githubusercontent.com · read 2026-10-31 |
| Microsoft Azure AI cloud | batch | $16.5per Mtok in$99per Mtok outraw.githubusercontent.com · read 2026-10-31 |
| Databricks cloud | standard | $30per Mtok in$180per Mtok out$30per Mtok cache writeraw.githubusercontent.com · read 2026-10-31 |
| DigitalOcean cloud | standard | $30per Mtok in$180per Mtok outmodels.dev · read 2026-10-31 |
| Microsoft Azure AI cloud | standard | $30per Mtok in$180per Mtok out$3per Mtok cachedmodels.dev · read 2026-10-31 |
| Microsoft Azure AI cloud | long-context batch | $30per Mtok in$135per Mtok outprices.azure.com · read 2026-08-24 |
| OpenAI api | standard | $30per Mtok in$180per Mtok outdevelopers.openai.com · read 2026-08-24 |
| Microsoft Azure AI cloud | long-context | $60per Mtok in$270per Mtok outprices.azure.com · read 2026-08-24 |
| Microsoft Azure AI cloud | priority | $60per Mtok in$360per Mtok out$6per Mtok cachedraw.githubusercontent.com · read 2026-10-02 |
| Nous Research 2 ways | $30per Mtok in$180per Mtok out | |
| Nous Research aggregator | batch | $15per Mtok in$90per Mtok outinference-api.nousresearch.com · read 2026-09-12 |
| Nous Research aggregator | standard | $30per Mtok in$180per Mtok outinference-api.nousresearch.com · read 2026-09-12 |
| OpenRouter 3 ways | $30per Mtok in$180per Mtok out | |
| OpenRouter aggregator | batch | $15per Mtok in$90per Mtok outopenrouter.ai · read 2026-08-24 |
| OpenRouter aggregator | openai/flex | $15per Mtok in$90per Mtok outopenrouter.ai · read 2026-08-25 |
| OpenRouter aggregator | standard | $30per Mtok in$180per Mtok outopenrouter.ai · read 2026-08-24 |
| Requesty aggregatornot on the seller's list today | azure | $30per Mtok in$180per Mtok out$30per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
| Requesty aggregator | openai | $30per Mtok in$180per Mtok out$30per Mtok cachedrouter.requesty.ai · read 2026-09-20 |
| Requesty aggregatornot on the seller's list today | openai-responses | $30per Mtok in$180per Mtok out$30per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
| Venice AI aggregator | standard | $30per Mtok in$180per Mtok outmodels.dev · read 2026-10-31 |
| Vercel AI Gateway aggregator | standard | $30per Mtok in$180per Mtok outai-gateway.vercel.sh · read 2026-08-25 |
Measured 18 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 4thof 101 | FrontierMath (Tiers 1-3) | mean_score (xhigh) | 50 |
| 5thof 46 | Enigma eval — Epoch AI | Accuracy | 23.82 |
| 6thof 56 | Humanity's Last Exam (Epoch AI replication) | Accuracy | 44.32 |
| 7thof 326 | GPQA Diamond | mean_score (xhigh) | 94.6 |
| 10thof 154 | Chess puzzles — Epoch AI | mean_score (xhigh) | 59 |
| 10thof 102 | Simplebench — Epoch AI | Score (AVG@5) | 0.741 |
| 15thof 273 | ARC-AGI-1 (Public Eval) | Score (%) (xhigh) | 98.25 |
| 15thof 122 | Frontiermath tiers 1 3 v2 — Epoch AI | mean_score (xhigh) | 82.456 |
| 16thof 260 | Critpt — Epoch AI | Accuracy (xhigh) | 30 |
| 24thof 78 | Frontiermath tier 4 v2 — Epoch AI | mean_score (xhigh) | 58.537 |
| 25thof 282 | ARC-AGI-2 (Public Eval) | Score (%) (xhigh) | 92.21 |
| 38thof 298 | ARC-AGI-2 (Semi-Private) | Score (%) (xhigh) | 83.33 |
| 39thof 93 | Simpleqa verified — Epoch AI | mean_score (xhigh) | 46.3 |
| 41stof 290 | Arc agi 2 — Epoch AI | Score (xhigh) | 83.33 |
| 42ndof 298 | ARC-AGI-1 (Semi-Private) | Score (%) (xhigh) | 94.5 |
| 43rdof 309 | Arc agi — Epoch AI | Score (xhigh) | 94.5 |
| 43rdof 111 | Vectara Hallucination Leaderboard | Hallucination Rate | 8.3 |
| 56thof 173 | Weirdml — Epoch AI | Accuracy (none) | 57.44 |
About
GPT-5.4 Pro — a text model from OpenAI, sold by 9 companies from $30 in and $180 out per million tokens, placed 7th of 326 on GPQA Diamond.
It takes text, images and files and returns text, with a context window of 1,050,000 tokens. It was published in March 2026, trained on material up to August 2025. Its sellers say it can reason step by step and call a tool. The catalogue files it under extract and reasoning. Nine companies sell it. The cheapest is $30 in and $180 out per million tokens at Databricks. Beside the standard rate there are batch, flex, separately routed and faster lanes. It has been measured on 18 boards, and stands best at 7th of 326 on GPQA Diamond.
Every current figure
- Maker
- OpenAI
- Register
- model
- Takes
- text + image + file
- Returns
- text
- Context
- 1,050,000 tokens
- Longest answer
- 128,000 tokens
- Published
- March 2026
- Knowledge to
- August 2025
- Licence
- not read
- Sellers
- 9
- Price
- $30 in and $180 out per million tokens — Databricks
- Boards
- 18
- Best place
- 7th of 326 — GPQA Diamond
Known as 10 names
GPT-5.4 Pro (XHigh)gpt-5.4-pro-2026-03-05_xhighgpt-5.4-pro-2026-03-05_unknown5.4 proopenai/gpt-5.4-proopenai/gpt-5.4-pro:batchgpt-5.4-proazure/openai-responses/gpt-5.4-pro@eastus2openai-responses/gpt-5.4-proOpenAI: GPT-5.4 Pro (batch)