OpenAI has introduced GPT-5.5, positioning it as their smartest and most intuitive model to date, representing the next step towards a new way of getting work done on a computer. GPT-5.5 is designed to understand user intent faster and can autonomously carry out more of the work itself.

The model demonstrates significant advancements in areas such as agentic coding, computer use, knowledge work, and early scientific research. It excels at planning, using tools, checking its work, navigating ambiguity, and persisting through complex, multi-part tasks.

GPT-5.5 matches GPT-5.4's per-token latency while delivering a much higher level of intelligence and using significantly fewer tokens to complete the same Codex tasks, enhancing its efficiency. It is also released with OpenAI's strongest set of safeguards to date, aimed at reducing misuse while preserving access for beneficial work.

In coding benchmarks, GPT-5.5 achieved 82.7% accuracy on Terminal-Bench 2.0 and 58.6% on SWE-Bench Pro, outperforming GPT-5.4 on Expert-SWE. For knowledge work, it scored 84.9% on GDPval and 78.7% on OSWorld-Verified, demonstrating enhanced capabilities across the full loop of information finding, understanding, tool use, and output generation.

In scientific research, GPT-5.5 shows clear improvements on GeneBench and BixBench, new evaluations focusing on multi-stage scientific data analysis in genetics and quantitative biology. Notably, an internal version of GPT-5.5 also contributed to discovering a new proof about Ramsey numbers in combinatorics.

Regarding inference efficiency, GPT-5.5 was co-designed for and trained on NVIDIA GB200 and GB300 NVL72 systems. Improvements in load balancing and partitioning heuristics, aided by Codex, led to over 20% faster token generation speeds. Cybersecurity capabilities are also advanced with stricter classifiers for potential cyber risk and the introduction of Trusted Access for Cyber for verified users to access advanced defensive features.

GPT-5.5 is rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, with API access coming soon. API pricing for gpt-5.5 is $5 per 1M input tokens and $30 per 1M output tokens, while gpt-5.5-pro is priced at $30 per 1M input tokens and $180 per 1M output tokens.

👉 Read the full article on OpenAI

I am sorry, but the content of this URL was inaccessible. Furthermore, I was unable to find alternative news articles or information within the specified tool operations.

👉 Read the full article on TechCrunch

I am sorry, but the content of this URL was inaccessible. Furthermore, I was unable to find alternative news articles or information within the specified tool operations.

👉 Read the full article on TechCrunch