OpenAI has introduced GPT-5.5, its newest flagship AI model, offering faster reasoning, stronger coding skills, and better performance on complex multi-step tasks across apps and tools.
The company says GPT-5.5 can understand user intent more quickly and work more independently. Instead of needing detailed instructions, the model can plan tasks, use tools, verify results, and continue working through unclear or complicated requests.
OpenAI says GPT-5.5 is particularly strong in software engineering, office productivity, scientific research, and computer automation.
GPT-5.5 is now rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex. GPT-5.5 Pro is also launching for Pro, Business, and Enterprise subscribers. API access is expected soon.
Stronger Coding Performance
One of the biggest upgrades in GPT-5.5 is coding. OpenAI says it performs better than GPT-5.4 while using fewer tokens, which may lower computing costs and reduce retries.
| Coding Benchmark | GPT-5.5 | GPT-5.4 |
| Terminal-Bench 2.0 | 82.7% | 75.1% |
| Expert-SWE (Internal) | 73.1% | 68.5% |
OpenAI says the model handles large codebases better, fixes unclear bugs more effectively, and can manage changes across multiple files.
Better for Workplace Tasks
GPT-5.5 is also designed for research, spreadsheets, writing documents, and using workplace software tools.
| Productivity Benchmark | GPT-5.5 | GPT-5.4 |
| GDPval | 84.9% | 83.0% |
| OSWorld-Verified | 78.7% | 75.0% |
OpenAI added that more than 85% of its own employees already use Codex weekly across teams such as engineering, finance, communications, marketing, and product management.
Improved Research and Math
OpenAI says GPT-5.5 also performs better in scientific and analytical tasks that require testing ideas, reviewing evidence, and reasoning through multiple steps.
| Research Benchmark | GPT-5.5 | GPT-5.4 |
| GeneBench | 25.0% | 19.0% |
| BixBench | 80.5% | 74.0% |
| FrontierMath Tier 4 | 35.4% | 27.1% |
Same Speed, Better Results
OpenAI says GPT-5.5 delivers stronger performance while maintaining similar speed to GPT-5.4. The model was trained and deployed on NVIDIA GB200 and GB300 NVL72 systems.
The company also said internal infrastructure improvements increased token generation speeds by more than 20%.
Pricing and Availability
In Codex, GPT-5.5 supports a 400K context window. The API version will launch soon with a 1 million token context window.
| API Pricing | Cost |
| Input Tokens | $5 per 1 million |
| Output Tokens | $30 per 1 million |
Safety Features
OpenAI says GPT-5.5 launches with its strongest safeguards yet, including stronger protections against cybersecurity misuse and biology-related risks.
With GPT-5.5, OpenAI is clearly focusing on AI systems that do more than answer questions — systems that can complete real work from start to finish.
