For months, Claude and ChatGPT have been locked in a stalemate — which LLM was the quickest, probably the most highly effective, one of the best? Up till now, many customers felt that Anthropic’s Claude Opus 4.7 had a slight edge in its bigger context window, versatile writing types, and superior information visualization, in addition to another key advantages. Effectively, that could be about to vary.
This week, OpenAI formally launched GPT-5.5, internally code-named “Spud”. However this new mannequin isn’t as potato-like as its nickname would indicate. In response to OpenAI’s official launch information, this mannequin is a basic redesign aimed toward “agentic” efficiency — the power for an AI to make use of a pc and plan forward, somewhat than merely reply to guide prompts. Will this be sufficient to convey customers again to ChatGPT after a mass exodus?
Associated
I moved my complete ChatGPT context to Claude and it lastly felt like residence
Right here is one of the best path to go from ChatGPT to Claude.
That is the mannequin that can shift individuals again to OpenAI
And it has a powerful beginning pitch
The last word objective of AI differs relying on who you ask, however most individuals agree that the dream is an AI that may deal with multi-step duties with out you holding its hand. That is the place GPT-5.5 beats Claude, arms down.
On Terminal-Bench 2.0, which checks advanced command-line workflows requiring planning, iteration, and gear coordination, GPT-5.5 achieved a staggering 82.7% accuracy. For comparability, Claude Opus 4.7 scored 69.4%, whereas Gemini 3.1 Professional scored 68.5%.
That’s an enormous 13% enchancment, however what does it really imply? That’s the distinction between an AI that means code, and an “agentic” AI — an AI that may really go into your system, run the terminal, and confirm the repair. This leap in efficiency is strictly why AI benchmarks nonetheless matter, even when they’re nonetheless usually wrapped up in overcomplicated “tech-bro” jargon.
However does this larger intelligence imply slower response instances? Apparently not: OpenAI claims it might probably match GPT-4’s per-token latency whereas being considerably smarter. It is usually extra “token environment friendly,” which means it usually makes use of fewer tokens to finish the identical activity as a result of it understands intent faster.
On the OSWorld-Verified benchmark, which measures how effectively AI can function a typical desktop OS, GPT-5.5 scored 78.7%, simply narrowly edging out Claude’s 78.0%.
OS
Android, iOS, Net
Developer
OpenAI
Value mannequin
Free with elective subscription
What’s the catch?
It received’t be low cost.
Within the OpenAI API, GPT-5.5 prices $5.00 per 1M enter tokens, which is double the value of the earlier technology. In order for you the much more exact GPT-5.5 Professional, designed for high-stakes environments like authorized analysis or information science, that worth jumps to $30.00.
API entry for GPT-5.5 is rolling out in phases to paying subscribers first. GPT‑5.5 is rolling out to Plus, Professional, Enterprise, and Enterprise customers in ChatGPT and Codex, and GPT‑5.5 Professional is rolling out to Professional, Enterprise, and Enterprise customers in ChatGPT from twenty third April, 2026. Will you be attempting it out?
