Here’s all you need to know:
Features of GPT-5.2
A few months after the rollout of GPT-5, OpenAI has introduced upgrades to the model. The company claims that the latest version is capable of creating spreadsheets and presentations, writing code, perceiving images, and understanding long contexts.
Long-context reasoning was tested on OpenAI’s MRCRv2 benchmark — an evaluation that measures a model’s ability to integrate information spread across long documents. This makes the model well-suited for deep analysis, synthesis, and complex multi-source workflows.
On image perception, the Sam Altman-led company said the model has a stronger grasp of visual elements, showing improved comprehension of images.
The company also noted that GPT-5.2 Thinking is more accurate than the GPT-5.1 Thinking model. The blog post added that GPT-5.2 sets new performancebenchmarks, outperforming industry professionals across 44 occupations.
GPT-5.2 plans, pricing
OpenAI has rolled out three variants of GPT-5.2 — Instant, Thinking, and Pro — all of which are available under paid plans. GPT‑5.2 is priced at $1.75/1 million input tokens and $14/1 million output tokens, with a 90% discount on cached inputs.
Performance benchmarks
GPT-5.2 outperformed human experts in knowledge-based tasks across 44 occupations. The company added that the Thinking plan of GPT-5.2 beat top industry professionals in 70.9% of comparisons on GDPval knowledge-work tasks. These tasks include making presentations, spreadsheets, and other work artefacts.
In terms of speed, GPT-5.2 Thinking produced outputs for given tasks at more than 11 times the speed and at less than 1% of the cost of expert professionals, the company said.
GPT-5.2 Thinking’s average score per task was 9.3 percentage points higher than GPT-5.1’s, increasing from 59.1% to 68.4%.
For coding, GPT-5.2 Thinking sets a new benchmarkwith a score of 55.6% on SWE-Bench Pro, which evaluates real-world software engineering tasks.
For professional use, GPT-5.2 Thinking can debug production code and implement complex features across large codebases.
How to access GPT-5.2?
To access OpenAI’s advanced AI model GPT-5.2, users must be subscribed to a Plus, Pro, Business, or Enterprise plan.
In ChatGPT, GPT-5.2 Instant handles everyday tasks, such as general assistance, technical writing, and translation.
GPT-5.2 Thinking helps with coding, summarising long documents, answering queries based on attached files, and solving math and logic problems.
The Pro version, which is yet to be fully explored, is expected to demonstrate stronger performance in highly complex tasks, such as advanced programming and system-level reasoning.
Heightened competition
Rivals, including Google and Microsoft, have been pumping billions of dollars into developing AI systems. Recently, OpenAI CEO Sam Altman had accepted that Google’s recent successes in AI could “create some temporary economic headwinds for our company,” The Information reported.
Google introduced Gemini 3, which the tech giant says is its most capable model yet. Software developers claim that it performs better in automating website and product design, as well as writing code.
Also Read: Rise of the machines: From AI to AGI to the uncharted realm of Superintelligence
Mark Zuckerberg is going all-in on a new high-stakes project: a “superintelligence” lab in Meta, focussed on building AI systems that surpass human-level intelligence.
Anthropic’s Claude unveiled Opus 4.5 in November, boosting Claude’s ability to write detailed code, create sophisticated agents, and streamline enterprise workflows through spreadsheet and financial analysis.