OpenAI Launches ChatGPT 5.5 With Stronger Coding and Computer-Use Skills to Take on Claude 4.7

OpenAI has unveiled GPT-5.5, its latest flagship AI model, promising faster reasoning, stronger coding performance, and better ability to complete multi-step tasks across apps and tools.

The company says GPT-5.5 is designed to understand user intent more quickly and take on more work independently. Instead of needing step-by-step instructions, the model can plan tasks, use tools, check results, and continue working through unclear or complex requests. OpenAI says the new release is especially strong in software engineering, office work, scientific research, and computer automation.

GPT-5.5 is rolling out now to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex. A more advanced GPT-5.5 Pro version is launching for Pro, Business, and Enterprise subscribers. OpenAI says API access is coming soon.

One of the biggest focuses of GPT-5.5 is coding. OpenAI says the model performs better than GPT-5.4 while using fewer tokens, which could lower computing costs and reduce retries.

On Terminal-Bench 2.0, a benchmark for complex command-line workflows, GPT-5.5 scored 82.7%, up from 75.1% for GPT-5.4. On OpenAI’s internal Expert-SWE test for long engineering tasks, GPT-5.5 reached 73.1%, compared with 68.5% for GPT-5.4.

OpenAI says the model is better at handling large codebases, debugging unclear issues, and carrying changes across multiple files.

Beyond coding, GPT-5.5 is built for knowledge work such as research, spreadsheet creation, document drafting, and navigating software tools.

On GDPval, a benchmark measuring professional task performance across 44 occupations, GPT-5.5 scored 84.9%, ahead of GPT-5.4 at 83.0%. On OSWorld-Verified, which tests whether models can operate computer environments on their own, GPT-5.5 scored 78.7%.

OpenAI says more than 85% of its own employees already use Codex weekly across teams including engineering, finance, communications, marketing, and product management.

OpenAI also says GPT-5.5 performs better in scientific workflows that require exploring ideas, analyzing evidence, and testing assumptions over multiple steps.

The company says the model scored 25.0% on GeneBench, up from 19.0% for GPT-5.4, and 80.5% on BixBench, compared with 74.0% for GPT-5.4.

On FrontierMath Tier 4, GPT-5.5 reached 35.4%, ahead of GPT-5.4 at 27.1%.

A common tradeoff with larger AI models is slower response times. OpenAI says GPT-5.5 avoids that problem by matching GPT-5.4 latency while delivering stronger performance.

The company says GPT-5.5 was trained and deployed on NVIDIA GB200 and GB300 NVL72 systems, and that internal infrastructure improvements increased token generation speeds by more than 20%.

OpenAI says GPT-5.5 launched with its strongest safeguards yet, including new protections for cybersecurity and biology-related misuse. The company says it tested the model internally and externally before release.

In Codex, GPT-5.5 supports a 400K context window. OpenAI says the API version will launch soon with a 1 million token context window, priced at $5 per 1 million input tokens and $30 per 1 million output tokens.

With GPT-5.5, OpenAI is clearly betting that users want AI systems that can do more than answer questions — they want software that can finish the job.