Model Tracker
GPT-5.2
GPT-5.2 is an AI model made by OpenAI. OpenAI released it on 11 December 2025. It is generally available. Its context window is 400,000 tokens.
In inglese
GPT-5.2 is an AI model made by OpenAI. OpenAI released it on 11 December 2025. It is generally available. Its context window is 400,000 tokens. Its longest answer is 128,000 tokens. Through the API it costs $1.75 per 1M tokens for input and $14.00 per 1M tokens for output. Its knowledge cutoff is 31 August 2025.
- Released
- 11 December 2025 — openai.com, read 19 October 2026
- Status
- Generally available — developers.openai.com, read 19 October 2026
- Takes
- Not confirmed: two readings disagree: text, image and text, image, code
- Returns
- Not confirmed: two readings disagree: text and text, image, code
- Open weights
- Not confirmed: OpenAI's pages reviewed do not state whether GPT-5.2 weights are open or downloadable.
- Licence
- Not confirmed: OpenAI's pages reviewed do not print a licence name for GPT-5.2.
- Context window
- 400,000 tokens — developers.openai.com, read 19 October 2026
- Longest answer
- 128,000 tokens — developers.openai.com, read 19 October 2026
- Knowledge cutoff
- 31 August 2025 — developers.openai.com, read 19 October 2026
- API price, input
- $1.75 per 1M tokens — developers.openai.com, read 19 October 2026
- API price, output
- $14.00 per 1M tokens — developers.openai.com, read 19 October 2026
- Where
- Not confirmed: two readings disagree: api and api, app
- Parameters
- Not confirmed: OpenAI's pages reviewed do not publish a parameter count for GPT-5.2.
- Model card
- https://cdn.openai.com/pdf/3a4153c8-c748-4b71-8e31-aecbde944f8d/oai_5_2_system-card.pdf — cdn.openai.com, read 19 October 2026
Benchmarks
Measured by others, not by the maker: each score with the name the measurer gives the model and the setting it ran, its day and its licence.
Arena
- Text Arena, gpt-5.2-high: 1,437 (1,434–1,441, 49,655 votes), published 15 October 2026
- Text Arena, gpt-5.2: 1,435 (1,432–1,439, 83,855 votes), published 15 October 2026
- Code Arena (WebDev), GPT-5.2: 1,417 (1,394–1,439, 1,081 votes), published 18 October 2026
- Vision Arena, gpt-5.2-high: 1,243 (1,237–1,249, 19,041 votes), published 9 October 2026
- Vision Arena, gpt-5.2: 1,229 (1,223–1,235, 20,127 votes), published 9 October 2026
Arena’s leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0: ratings from people’s votes between two anonymous models, as Arena publishes them. Data · CC BY 4.0
Epoch AI
- Epoch Capabilities Index, GPT-5.2: 153.5 (151.6–155.3), as of 19 October 2026
- GPQA Diamond, gpt-5.2-2025-12-11_xhigh: 91.4% (± 1.8), run 13 December 2025
- GPQA Diamond, gpt-5.2-2025-12-11_high: 88.2% (± 1.9), run 11 December 2025
- GPQA Diamond, gpt-5.2-2025-12-11_medium: 87.9% (± 1.9), run 11 December 2025
- GPQA Diamond, gpt-5.2-2025-12-11_low: 82.7% (± 2.3), run 11 December 2025
- GPQA Diamond, gpt-5.2-2025-12-11_none: 73.2% (± 3.2), run 13 July 2026
- FrontierMath, tiers 1–3, gpt-5.2-2025-12-11_xhigh: 67.4% (± 2.8), run 11 June 2026
- FrontierMath, tier 4, gpt-5.2-2025-12-11_xhigh: 31.7% (± 7.3), run 11 June 2026
- SWE-bench Verified, gpt-5.2-2025-12-11_high: 73.8% (± 2.0), run 12 February 2026
- OTIS Mock AIME 2024–2025, gpt-5.2-2025-12-11_high: 96.1% (± 2.6), run 11 December 2025
- OTIS Mock AIME 2024–2025, gpt-5.2-2025-12-11_xhigh: 96.1% (± 2.7), run 13 December 2025
- OTIS Mock AIME 2024–2025, gpt-5.2-2025-12-11_medium: 93.9% (± 3.1), run 11 December 2025
- OTIS Mock AIME 2024–2025, gpt-5.2-2025-12-11_low: 78.9% (± 5.0), run 11 December 2025
- OTIS Mock AIME 2024–2025, gpt-5.2-2025-12-11_none: 62.2% (± 7.3), run 13 July 2026
- SimpleQA Verified, gpt-5.2-2025-12-11_xhigh: 37.1% (± 1.5), run 27 August 2026
- SimpleQA Verified, gpt-5.2-2025-12-11_high: 34.3% (± 1.5), run 27 August 2026
- SimpleQA Verified, gpt-5.2-2025-12-11_low: 32.8% (± 1.5), run 27 August 2026
- SimpleQA Verified, gpt-5.2-2025-12-11_medium: 32.7% (± 1.5), run 27 August 2026
Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks, CC BY 4.0: Epoch AI’s own runs and its Capabilities Index. Data · CC BY 4.0
Questions
When was GPT-5.2 released?
OpenAI released GPT-5.2 on 11 December 2025.
Is GPT-5.2 still available?
Yes: it is generally available, as OpenAI states it.
What is the context window of GPT-5.2?
The context window of GPT-5.2 is 400,000 tokens, as OpenAI prints it.
How much does the GPT-5.2 API cost?
On the standard tier, GPT-5.2 costs $1.75 per 1M tokens for input and $14.00 per 1M tokens for output, as OpenAI lists it.
What is the knowledge cutoff of GPT-5.2?
The knowledge cutoff of GPT-5.2 is 31 August 2025.
More from OpenAI
- ada
- babbage
- babbage-002
- code-davinci-002
- computer-use-preview
- curie
- dall-e-3
- DALL·E
- DALL·E 2
- davinci
- davinci-002
- GPT-3.5 Turbo
- GPT-3.5 Turbo Instruct
- GPT-4
- GPT-4 Turbo
- GPT-4.1
- GPT-4.1 mini
- GPT-4.1 nano
- GPT-4.5
- GPT-4.5 Preview
- GPT-4o
- GPT-4o mini
- GPT-4o Mini Search Preview
- GPT-4o Mini Transcribe
Sources
- All models | OpenAI API (read 19 October 2026)
- GPT-5.2 Model | OpenAI API (read 19 October 2026)
- Introducing GPT-5.2 (read 19 October 2026)
- Deprecations | OpenAI API (read 19 October 2026)
- Update to GPT-5 System Card: GPT-5.2 (read 19 October 2026)