OpenAI Releases GPT-5.5 (Spud), Its Most Capable Model Yet

As rumored earlier, OpenAI has released GPT-5.5, codenamed “Spud.” The company says it’s a major upgrade over GPT-5.4. The latest model is now rolling out across ChatGPT and Codex, with enterprise and API access coming soon.

OpenAI is positioning GPT-5.5 toward more autonomous AI, where systems don’t just respond but actively complete full tasks. That includes coding, research, and tool-based workflows with less human guidance.

Benchmark results show GPT-5.5 is a clear step up

On performance benchmarks, GPT-5.5 consistently comes out ahead of GPT-5.4 across most categories. In coding-heavy evaluations, it shows stronger results in multi-step problem solving, terminal-based tasks, and real-world software engineering challenges. The model beats Anthropic’s Claude Opus 4.7 and Gemini 3.1 Pro on all major benchmarks.

In reasoning and professional workloads, GPT-5.5 also edges forward, especially in document-heavy and tool-using environments where models need to stay consistent across longer workflows. It is also reported to handle computer-use tasks more reliably, where earlier versions often struggled with continuity.

Across knowledge work simulations, GPT-5.5 shows gains in tasks that mirror real office environments like data analysis, reporting, and workflow automation. The improvements are not just about accuracy but also efficiency, with fewer retries and reduced token usage for the same tasks.

Efficiency and real-world performance shift

One of the key changes is speed consistency. OpenAI says GPT-5.5 matches GPT-5.4 latency in real use while delivering higher-quality outputs. That balance between speed and capability is becoming central as models grow more complex. The model also shows stronger efficiency in coding workflows, as it completes tasks with fewer tokens while maintaining output quality. That being said, benchmark gains don’t always translate perfectly into messy real-world environments.

Availability and pricing

OpenAI is rolling out GPT‑5.5 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. In ChatGPT, a dedicated “Thinking” mode is available for deeper problem solving, while GPT-5.5 Pro helps with more demanding workflows that require higher precision and reliability.

In Codex, the model is available across multiple plans including Plus, Pro, Business, Enterprise, Edu, and Go, with support for a large 400K context window. A faster mode is also included, which boosts generation speed at a higher usage cost.

For developers, GPT-5.5 will soon be released in the API through both Responses and Chat Completions endpoints. OpenAI will charge $5 per 1M input tokens and $30 per 1M output tokens, with a 1M context window.