OpenAI has unveiled its latest language model, GPT-5.5, marking a significant advancement in artificial intelligence capabilities. This new model is engineered for sophisticated computer interactions, capable of writing and debugging code, browsing the web, managing spreadsheets, and executing complex multi-step tasks with minimal human oversight. The rollout commenced today for subscribers of ChatGPT Plus, Pro, Business, and Enterprise tiers, with API access expected shortly.
Key Takeaways
- GPT-5.5 is now available for paid ChatGPT and Codex users, with API access forthcoming.
- The model demonstrates superior performance across various benchmarks, particularly in complex command-line workflows and general knowledge work.
- Despite a higher per-token cost, GPT-5.5 achieves greater efficiency by using fewer tokens for equivalent tasks compared to its predecessor, GPT-5.4.
- API pricing is set at $5 per million input tokens and $30 per million output tokens for the standard version, with a premium tier for GPT-5.5 Pro.
OpenAI’s announcement highlights GPT-5.5 as their most intelligent and user-friendly model to date, representing a new paradigm for computer-based work. The improvements are particularly pronounced in areas requiring advanced reasoning, tool utilization, and sustained action over time, such as agentic coding, knowledge work, and early-stage scientific research.
Introducing GPT-5.5
A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done.
Now available in ChatGPT and Codex.
A notable achievement for GPT-5.5 is its ability to match the performance speed of GPT-5.4 while simultaneously surpassing it on nearly all evaluated benchmarks. This leap in efficiency, where a more capable model does not incur additional latency on the same hardware, is a rare and significant engineering feat. Traditionally, larger and more complex models tend to require more processing time.
On the Terminal-Bench 2.0, which assesses a model’s proficiency in executing intricate command-line operations requiring planning and iterative tool use, GPT-5.5 achieved an impressive 82.7%. This figure significantly outpaces competitors like Claude Opus 4.7 (69.4%) and Gemini 3.1 Pro (68.5%). Furthermore, the GDPval benchmark, designed to evaluate knowledge work across 44 diverse professional domains, shows GPT-5.5 meeting or exceeding human professional performance in 84.9% of comparisons.
Regarding coding capabilities, GPT-5.5 also shows marked improvement. On Expert-SWE, a benchmark for long-duration coding tasks, it surpasses GPT-5.4. While Claude Opus 4.7 leads on SWE-Bench Pro, OpenAI suggests this may be due to potential memorization issues by the competitor on certain problems. This rapid development cycle, with new models emerging frequently, reflects the intense pace of innovation in the AI sector.
The availability of GPT-5.5 is currently limited to paid subscribers, excluding free-tier users. This tiered access model is common for advanced AI services, reflecting the substantial computational resources required to develop and deploy such sophisticated models.
The integration of GPT-5.5 into OpenAI’s Codex platform is particularly noteworthy. Codex, designed as an agentic coding environment, benefits significantly from GPT-5.5’s advanced reasoning and task execution capabilities. Users are reporting a sensation akin to collaborating with a more intelligent entity, suggesting a profound shift in developer productivity.
GPT-5.5 Pro, tailored for more demanding applications, achieves a 90.1% score on BrowseComp, a benchmark measuring the ability to retrieve obscure information online, exceeding Gemini 3.1 Pro’s 85.9%. The model also ranks highest on the Artificial Analysis Index, indicating its overall intelligence and efficiency in processing and generating information.
The API pricing introduces a shift, with GPT-5.5 set at $5 per million input tokens and $30 per million output tokens, and GPT-5.5 Pro at $30 and $180 respectively. While these rates are higher than GPT-5.4’s, OpenAI CEO Sam Altman emphasizes that the enhanced token efficiency of GPT-5.5 makes it more cost-effective for equivalent tasks. This improved efficiency is crucial for advanced AI applications, especially in areas like blockchain development and Web3 infrastructure, where optimizing resource usage is paramount.
The comparative pricing against other leading AI models, such as Xiaomi MiMo v2.5 Pro, Minimax M2.7, and Kimi K2.5, positions GPT-5.5 competitively, especially when factoring in its superior performance and efficiency gains. This release signifies a continued push towards more capable and autonomous AI systems, with implications for how AI integrates with decentralized technologies and contributes to the evolving Web3 landscape.
Based on materials from : decrypt.co