英文互译镜像站

GPT-5.2

Last updated

GPT-5.2
Developer OpenAI
Initial releaseDecember 11, 2025;2 months ago (2025-12-11)
Included with
Predecessor GPT-5.1
Successor GPT-5.3
Type
License Proprietary
Website openai.com/index/introducing-gpt-5-2/

GPT-5.2 is OpenAI's latest flagship large language model, released on December 11, 2025. [1] Succeeding GPT-5.1, it is a family of three large language models within the GPT series. It comes in two modes: GPT-5.2 instant and GPT-5.2 thinking (with standard and extended thinking), with the latter being a reasoning model. A GPT-5.2 Pro model was released on the same day, which takes more reasoning time and compute than GPT-5.2 thinking. [1] OpenAI also released a variant specialized for coding, GPT-5.2-Codex. [2]

Contents

Development and release

The release of GPT-5.2 occurred on December 11, 2025, approximately three weeks after the release of Google's Gemini 3 Pro, which had been widely cited as the leading multimodal model at the time. [3] Media outlets reported that an internal "Code Red" memo, [4] [5] prompted by Gemini's dominance, was a key factor in accelerating the model's release from an originally planned late-December window. [1] OpenAI executives disputed the narrative that the launch was solely reactive. Speaking to Wired , Fidji Simo, OpenAI's CEO of Applications, stated that the company had "been working on this model’s release for months", though she acknowledged that the "Code Red" and additional resources allocated to ChatGPT was "helpful" in finalizing the deployment. [6]

Architecture and capabilities

GPT-5.2 is better at spreadsheet creation, financial modeling, presentations, and multi-step project execution. [7]

GPT‑5.2 comes in three modes; GPT-5.2 instant, GPT-5.2 thinking and GPT-5.2 Pro. [8] Instant model is designed for speed and efficiency, and the other two models are reasoning models. GPT-5.2's Pro version comes into two separate selectors: standard and extended. These two modes use different compute times and powers.

According to METR GPT-5.2 (high) has a 50%-time horizon of 6 hours 34 minutes and a 80%-time horizon of 55 minutes. This made GPT-5.2-high the model with the longest task-completion time horizon until 20 February 2026, when it was overtaken by Claude Opus 4.6. [9]

Reception

VentureBeat noted that early reactions suggest that GPT-5.2 was "used for power users, developers, and enterprise agents rather than casual chat". Aaron Levie, CEO of Box, reported that the model scores "7 points better than GPT-5.1" on the company's tests for real-world knowledge work. [10]

The Guardian reported GPT-5.2 using Elon Musk's encyclopedia Grokipedia as a source, and Grokipedia itself was criticized by Nina Jankowicz as poorly sourced. [11]

References

  1. 1 2 3 Field, Hayden (December 11, 2025). "GPT-5.2 is OpenAI's latest move in the agentic AI battle". The Verge. Archived from the original on December 13, 2025. Retrieved December 13, 2025.
  2. ↑ "Enterprise AI coding grows teeth: GPT‑5.2‑Codex weaves security into large-scale software refactors". VentureBeat. December 18, 2025.
  3. ↑ Bellan, Rebecca (December 11, 2025). "OpenAI fires back at Google with GPT-5.2 after 'code red' memo". TechCrunch. Archived from the original on December 16, 2025. Retrieved December 16, 2025.
  4. ↑ "Anthropic CEO Dario Amodei makes fun of 'Code Red' at OpenAI and Google; says: Felt no need to ..." The Times of India. December 4, 2025. ISSN   0971-8257 . Retrieved December 16, 2025.
  5. ↑ Warren, Tom (December 5, 2025). "OpenAI's GPT-5.2 'code red' response to Google is coming next week". The Verge. Retrieved December 16, 2025.
  6. ↑ Zeff, Maxwell. "OpenAI Launches GPT-5.2 as It Navigates 'Code Red'". Wired. ISSN   1059-1028. Archived from the original on December 16, 2025. Retrieved December 16, 2025.
  7. ↑ "OpenAI launches GPT-5.2 after 'code red' push to counter Google's Gemini 3". Reuters. December 11, 2025. Archived from the original on December 30, 2025. Retrieved December 14, 2025.
  8. ↑ Capoot, Ashley (December 11, 2025). "Sam Altman expects OpenAI to exit 'code red' by January after launch of GPT-5.2 model". CNBC. Archived from the original on December 11, 2025. Retrieved December 16, 2025.
  9. ↑ "Task-Completion Time Horizons of Frontier AI Models". METR . February 20, 2026. Retrieved February 20, 2026.
  10. ↑ "GPT-5.2 first impressions: a powerful update, especially for business tasks and workflows". VentureBeat. December 11, 2025.
  11. ↑ Down, Aisha (January 24, 2026). "Latest ChatGPT model uses Elon Musk's Grokipedia as source, tests reveal". The Guardian. ISSN   0261-3077 . Retrieved February 13, 2026.
This page is based on this Wikipedia article
Text is available under the CC BY-SA 4.0 license; additional terms may apply.
Images, videos and audio are available under their respective licenses.
霸屏SEO镜像站群 量子镜像站群 站群镜像程序 泛域名镜像 站群克隆软件