OpenAI Launches GPT-5.2 After Internal “Code Red” Over Google’s Gemini Threat

Udaipur, Dec 12, 2025: OpenAI has officially unveiled GPT-5.2, the latest addition to its AI model lineup for ChatGPT, released in three variants: Instant, Thinking, and Pro. The launch follows an internal “code red” directive issued earlier this month by CEO Sam Altman, urging rapid improvements to ChatGPT amid growing competition from Google’s Gemini 3.

During a press briefing, OpenAI chief product officer Fidji Simo said GPT-5.2 is designed to “unlock even more economic value for people,” highlighting its upgraded capabilities in spreadsheet automation, presentation creation, coding, image understanding, long-context processing, tool usage, and complex multi-step tasks.

Three Models, Different Purposes

• Instant: Optimized for speed—ideal for writing, translation, and quick tasks.
• Thinking: Produces structured “thinking” traces to handle advanced reasoning, coding, and mathematics.
• Pro: A higher-accuracy version with deeper simulated reasoning designed for the most challenging workloads.

GPT-5.2 now supports a 400,000-token context window, enabling it to process hundreds of documents simultaneously. The model’s knowledge cutoff is August 31, 2025.

The rollout begins today for paid ChatGPT subscribers, with API access also available. Pricing is set at $1.75 per million input tokens, a 40% increase over GPT-5.1. The older GPT-5.1 will remain accessible for three months under a legacy models menu.

Playing Catch-up with Google

OpenAI’s release comes during a turbulent period. After Google’s Gemini 3 outperformed rivals on several benchmarks and surged in adoption, Altman issued a company-wide “code red.” Teams were directed to prioritize improvements to ChatGPT, even delaying advertising plans.

The pressure is significant: OpenAI has committed $1.4 trillion to AI infrastructure over the coming years. Meanwhile, Google’s Gemini app boasts 650 million monthly active users, and ChatGPT reports 800 million weekly active users.

GPT-5.2 marks OpenAI’s third major model update since August. GPT-5 introduced mode routing but drew complaints about detached, clinical responses. GPT-5.1 followed with new personality presets and a more conversational tone.

Benchmarks: Strong Gains, Limited Comparisons to Gemini

Oddly, OpenAI avoided direct comparisons with Gemini 3 on its official site, focusing instead on internal benchmarks and improvements over earlier GPT-5 models. However, during the press briefing, it shared selective comparisons:

• SWE-Bench Pro (Software Engineering):

  • GPT-5.2 Thinking: 55.6%
  • Gemini 3 Pro: 43.3%
  • Claude Opus 4.5: 52.0%

• GPQA Diamond (graduate-level Science):

  • GPT-5.2: 92.4%
  • Gemini 3 Pro: 91.9%

OpenAI reports that GPT-5.2 Thinking matches or surpasses human professionals on 70.9% of tasks in its new GDPval benchmark, compared to 53.3% for Gemini 3 Pro.

The company also claims the model completes tasks 11× faster and at under 1% of the cost of human experts.

According to Max Schwarzer, OpenAI’s post-training lead, GPT-5.2 generates 38% fewer hallucinations than GPT-5.1—delivering far more reliable results.

As always, experts caution that benchmarks can be framed to favor a company’s narrative. Independent evaluations will take time to surface, and real-world performance varies by task and user.

For now, GPT-5.2 appears to offer incremental but meaningful improvements, especially for professional work, coding, and long-context reasoning—while setting the stage for the escalating AI rivalry between OpenAI and Google.

 #OpenAI #GPT52 #ChatGPT #AIUpdate #Gemini3 #TechNews #AIModels #UdaipurNews #RajasthanUpdates #Innovation2025 #ArtificialIntelligence