GPT-5.2vsGPT-5.4
Both language models, compared at each maker’s own API list prices. Every figure links to its source.
Compare at your usage
Every model priced for the same request8,000 tokens in, 2,000 out
| GPT-5.2 on OpenAI API | GPT-5.4 on OpenAI API | |
|---|---|---|
| Rate applied | Base rate | Base rate |
| Inputper 1M tokens | $1.75 | $2.50 |
| Cache readper 1M tokens | $0.175 | $0.25 |
| Outputper 1M tokens | $14.00 | $15.00 |
| Cost per request | $0.042 | $0.050 |
| Per 1,000 requests | $42.00 | $50.00 |
- GPT-5.2 OpenAI API pricing · as of Nov 2, 2026
- GPT-5.4 OpenAI API pricing · as of Nov 2, 2026
Every model is priced for the same request at its maker’s own list prices (each with its source and date), so bold marks the cheapest. A model that can’t serve the request is named, not ranked; a condition its seller doesn’t price says which rate was used.
Cost per request as prompts grow
2K tokens out, list price on each maker’s API. GPT-5.4 charges $5.00 / $22.50 above 272K tokens. A line ends at the model’s context window.
- GPT-5.2 · OpenAI API
- GPT-5.4 · OpenAI API
Where each stands among all scored models
Each line runs from the lowest to the highest score on the board; the tick is the median.
- GPT-5.2
- GPT-5.4
Key differences
These differences use verified vendor specifications or each maker’s official API on the same region, currency, meter and billing conditions. Different conditions remain side by side in the full table; they do not establish which model is cheaper. Declared features describe support, not measured quality or a recommendation. Missing evidence does not mean a feature is unsupported.
Full comparison
| GPT-5.2 OpenAI | GPT-5.4 OpenAI | |
|---|---|---|
| Overview | ||
| Released | Dec 11, 2025 | Mar 5, 2026 |
| Open weights | — | — |
| Knowledge cutoff | Aug 2025 | Aug 2025 |
| Specs | ||
| Context | 400K | 1.05M |
| Max input | 272K | 922K |
| Max output | 128K | 128K |
| Input modalities | Text, Image | Text, Image |
| Output modalities | Text | Text |
| Reasoning | ||
| Reasoning efforts | none, low, medium, high, xhigh | none, low, medium, high, xhigh |
| Reasoning can be off | ||
| Tool use | ||
| Structured output | ||
| Pricing · per 1M tokens | ||
| Input price | $1.75 | $2.50 |
| Output price | $14.00 | $15.00 |
| Cache read price | $0.175 | $0.25 |
| Blended price (3 input : 1 output) | $4.8125 | $5.625 |
| Pricing under conditions · per 1M tokens | ||
| Input price, long prompts | Not offered | $5.00Prompts over 272K tokens |
| Output price, long prompts | Not offered | $22.50Prompts over 272K tokens |
| Batch input price | $0.875Batch | $1.25Batch |
| Batch output price | $7.00Batch | $7.50Batch |
| Flex tier input price | $0.875Flex tier | $1.25Flex tier |
| Flex tier output price | $7.00Flex tier | $7.50Flex tier |
| Cited scores | ||
| Epoch Capabilities Index | 154.4 | 157.8 |
| LMArena Text | 1,437 high effort | 1,475 high effort |
| LMArena Text · Coding | 1,489 high effort | 1,520 high effort |
| LMArena Vision | 1,246 high effort | 1,303 high effort |
| LMArena Text · Chinese | 1,464 high effort | 1,506 high effort |
| Epoch AI · GPQA Diamond | 91.4% xhigh effort | 93.3% xhigh effort |
| Epoch AI · FrontierMath Tiers 1–3 | 67.4% xhigh effort | 78.6% xhigh effort |
| Epoch AI · SWE-bench Verified | 73.8% high effort | 76.9% high effort |
Headline prices may come from different channels, regions or tiers. Check each source before choosing a model; missing facts stay blank.
Model details and sources
Cited scores
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · Epoch AI Benchmarking Hub data · CC BY 4.0 · as of Nov 5, 2026
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · Epoch AI Benchmarking Hub data · CC BY 4.0 · as of Oct 28, 2026
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · Epoch AI Benchmarking Hub data · CC BY 4.0 · as of Nov 4, 2026
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · Epoch AI Benchmarking Hub data · CC BY 4.0 · as of Oct 7, 2026
- EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · EQ-Bench: Longform Writing · MIT · as of Oct 28, 2026
- LiveBench (White et al.), release 2026-06-25, livebench.ai, licensed CC BY-SA 4.0. This board, adapted from it, is shared under CC BY-SA 4.0. · LiveBench release 2026-06-25 · CC BY-SA 4.0 · as of Jun 25, 2026
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · LMArena leaderboard dataset (Hugging Face) · CC BY 4.0 · as of Sep 13, 2026
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · LMArena leaderboard dataset (Hugging Face) · CC BY 4.0 · as of Oct 28, 2026
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · LMArena leaderboard dataset (Hugging Face) · CC BY 4.0 · as of Oct 19, 2026
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · Epoch AI Benchmarking Hub data · CC BY 4.0 · as of Nov 2, 2026
- EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · EQ-Bench: 4 · MIT · as of Oct 24, 2026