OpenAI is pushing AI past chat with the current launch of GPT-5.5, a mannequin designed to finish multi-step work as an alternative of stopping at solutions.
The corporate launched GPT-5.5 on April 23, a brand new flagship AI mannequin designed to deal with multi-step duties throughout software program, analysis, and on a regular basis laptop work. It strikes towards agentic techniques that plan, act, and full jobs with minimal steering.
OpenAI claims that GPT-5.5 can deal with loosely outlined requests by breaking them into steps. It could possibly use instruments, confirm outcomes, and proceed working till the duty is full.
The corporate rolled GPT-5.5 out throughout ChatGPT and Codex for Plus, Professional, Enterprise, and Enterprise customers, with API entry anticipated later. A better-accuracy GPT-5.5 Professional tier targets extra demanding workloads.
AI fashions are shifting from solutions to accomplished work
OpenAI factors to a spread of benchmark outcomes to indicate how GPT-5.5 improves on earlier fashions. In coding checks that replicate actual growth work, GPT-5.5 scores 82.7% in comparison with 75.1% for GPT-5.4, exhibiting it could full extra duties from begin to end.
GPT-5.5 scores 82.7% in comparison with 75.1% for GPT-5.4. Picture credit score: OpenAI
GPT-5.5 performs higher in broader checks, dealing with complicated information work and working software program instruments, suggesting reliability throughout longer, multi-step duties. The mannequin completes extra duties end-to-end, not like earlier variations that stopped midway.
OpenAI pairs these modifications with enhancements in effectivity. GPT-5.5 matches GPT-5.4’s response velocity whereas utilizing fewer tokens to finish the identical duties.
Decrease token utilization reduces back-and-forth and shortens the trail to a completed consequence, which issues for each value and value. A mannequin that reaches a solution with fewer steps tends to supply extra predictable outcomes.
OpenAI is pushing AI deeper into on a regular basis laptop use
OpenAI frames GPT-5.5 as a device for information work past coding. It generates paperwork, builds spreadsheets, analyzes knowledge, and transitions between apps.
Instance of an app coded with ChatGPT/Codex. Picture credit score: OpenAI
Examples embody automating enterprise reporting, processing monetary paperwork, and managing operational requests.
GPT-5.5 can help in multi-step problem-solving, from info gathering to concept testing and consequence refinement. Whereas this functionality continues to be in its early phases, it suggests AI might more and more deal with the iterative work of analysis.
ChatGPT’s pricing ranges from a free tier with restricted entry to higher-end plans beginning at $8 per thirty days for Go and $20 per thirty days for Plus. Professional plans begin at $100 per thirty days for expanded utilization and entry to GPT-5.5 Professional.
Enterprise plans value $20 per consumer per thirty days. Enterprise pricing is customizable, with larger tiers providing extra entry, compute, and superior options like expanded reminiscence, deep analysis, and agent capabilities.