ChatGPT-maker OpenAI has unveiled the latest version of its artificial intelligence model — GPT-5.5. The company claims that the new model comes with significant improvements in reasoning, coding and scientific research. The company has also stated that GPT-5.5 is faster, more efficient and safer as compared to its predecessor. GPT‑5.5 builds on the capabilities of GPT‑5.4 but introduces key upgrades. The latest model can easily handle complex multi-step tasks with better planning, tool use and self-correction. The model is particularly strong in coding and debugging. OpenAI also stated that it managed to achieve 82.7% accuracy on Terminal‑Bench 2.0, outperforming rivals like Anthropic’s Claude Opus 4.7 and Google’s Gemini 3.1 Pro.Despite its increased intelligence, GPT‑5.5 maintains the same speed as GPT‑5.4 while using fewer tokens per task, making it more cost‑effective for enterprises and developers.
New variants for different needs
OpenAI is offering GPT‑5.5 in multiple versions:
- GPT‑5.5 Thinking: Optimized for faster, smarter to complex problems.
- GPT‑5.5 Pro: Designed for higher‑quality outputs with reduced latency, ideal for demanding workflows.
These variants aim to give businesses and professionals flexibility depending on their use cases, from research to enterprise automation.
Safety and risk management
OpenAI also noted that GPT-5.5 comes with its strongest safeguards yet. The model has undergone extensive red-dead testing to access cybersecurity and biology risks, with input from nearly 200 early-access partners. It is classified as “High” risk but below the “Critical” threshold, meaning it could amplify existing pathways to harm but does not introduce unprecedented new risks. This cautious rollout reflects growing scrutiny of advanced AI systems and their potential misuse.GPT‑5.5 is rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, and GPT‑5.5 Pro is rolling out to Pro, Business, and Enterprise users in ChatGPT. “API deployments require different safeguards and we are working closely with partners and customers on the safety and security requirements for serving it at scale. We’ll bring GPT‑5.5 and GPT‑5.5 Pro to the API very soon,” said the company.
What’s new compared to GPT‑5.4
| Feature | GPT‑5.4 | GPT‑5.5 | Improvement |
| coding | strong | State‑of‑the‑art agentic coding | More accurate, fewer retries |
| Context Handling | Limited | Better reasoning across large systems | Handles ambiguity more effectively |
| speed | fast | Same latency | Maintains speed despite larger model |
| efficiency | Higher token usage | Fewer tokens per task | Lower compute costs |
| Safety | standard safeguards | Strongest safeguards yet | Expanded cyber & bio risk testing |
