IT Brief US - Technology news for CIOs & IT decision-makers
United States
OpenAI launches GPT-6.2 models for work & office tasks

OpenAI launches GPT-6.2 models for work & office tasks

Tue, 17th Nov 2026 (Today)
Mara Sugue
MARA SUGUE News Editor

OpenAI has launched two new models, GPT-6.2 Astra and GPT-6.2 Sol, aimed at complex professional work and everyday office tasks.

GPT-6.2 Astra is positioned as the stronger of the two for work such as financial modelling, legal analysis and scientific research. GPT-6.2 Sol is intended for multi-step tasks including writing, coding and project planning.

Both models will roll out in ChatGPT Work and Codex, and will also be available through the API. OpenAI said they are designed to produce stronger results while using fewer output tokens, which could lower the cost of each task for customers working within usage limits.

Benchmark results

OpenAI tied the launch to a set of benchmark comparisons against its earlier systems and rival models. On Agents' Last Exam, which measures multi-step professional workflows, GPT-6.2 Astra scored 5.2 points higher than GPT-6 Astra at less than one-third of the cost, according to the company.

On the same test, Astra scored 5.5 points ahead of Claude Fable 5.2 at 5% of the cost, OpenAI said. It added that the published leaderboard results for Fable 5.2 excluded nine tasks that used a fallback model.

For scientific work, Astra reached new highs on Terminal-Bench-Science 0.1 and the Vals Science Index, according to OpenAI. Terminal-Bench-Science 0.1 tests whether a model can carry out scientific work in a terminal, and Astra scored nine percentage points higher than GPT-6 Astra at less than one-third of the cost.

OpenAI also highlighted legal and business workflow tests. On Harvey LAB-AA v1.1, which uses legal assignments to assess whether a model can identify relevant facts and produce an accurate deliverable, GPT-6.2 Astra more than doubled Claude Fable 5.2's pass rate at about one-sixth of the cost, the company said.

On GDPval v2, an internal benchmark built around long-horizon assignments from consulting, finance, design and other sectors, Astra improved by 6.2 percentage points over GPT-6 Astra while using 22% fewer output tokens, OpenAI said. GPT-6.2 Sol improved by 5.3 points over GPT-6.1 Sol on the same benchmark, according to the company.

Everyday model

OpenAI described GPT-6.2 Sol as its standard model for routine work. It is intended for tasks that take several steps, such as converting meeting notes into project plans, tracing software bugs and handling everyday personal organisation tasks.

In Cognition's FrontierCode 2.0 benchmark, GPT-6.2 Sol came within 0.5 points of GPT-6.2 Astra at about one-fifth of the cost when both were run at maximum settings, OpenAI said. FrontierCode 2.0 tests whether changes to real codebases meet maintainers' standards for merging.

OpenAI also pointed to its internal LifeVal evaluation, which measures performance on difficult everyday tasks tailored to individual needs. GPT-6.2 Sol outscored GPT-6 Astra at about one-seventh of the cost, according to the company.

Examples cited by OpenAI included shortlisting kindergartens after a move, finding a home that fits a family's budget and commute, and planning meals for a parent adapting to a new diagnosis.

Pricing changes

The launch also brings lower API pricing for Astra than the previous version. OpenAI said GPT-6.2 Astra's input and output token prices are 20% lower than GPT-6 Astra's, with a 60% discount on cached input.

It listed the pricing at USD $8 for input, USD $0.40 for cached input and USD $40 for output per million tokens. GPT-6.2 Sol keeps the same API token prices as GPT-6.1 Sol, OpenAI said.

Cost is central to the launch. OpenAI said several benchmark results showed the new models close to or ahead of other leading systems at a lower cost per task, reflecting broader competition over balancing model quality with operating expense.

Professional output

OpenAI said it trained GPT-6.2 Astra to produce slides, spreadsheets and documents that better match the conventions used in fields including law, sales, finance and accounting. It said the model's writing is more direct and specific, with less filler and fewer generic conclusions.

OpenAI also cited results from Artificial Analysis's AA-AnalystAgent benchmark, which tests work on real spreadsheets and reports. In one task using California Medicaid expenditure reports, GPT-6.2 Astra answered 62.5% of questions correctly across five independent attempts, according to the company.

The launch underscores how model developers are increasingly targeting practical office and professional workflows rather than only general-purpose chatbot use. OpenAI said GPT-6.2 Astra is its model for harder projects, while GPT-6.2 Sol is intended for a broader set of daily analytical, writing and coding tasks.