OpenAI just released GPT-5.5, calling it “a new class of intelligence for real work.” After digging into the announcement, here’s what stands out.
Agentic Coding Gets a Major Upgrade#
GPT-5.5 scores 82.7% on Terminal-Bench 2.0, up from GPT-5.4’s 75.1%. On Expert-SWE (OpenAI’s internal eval for long-horizon coding tasks with ~20-hour median human completion time), it hits 73.1% vs 68.5%. The key improvement isn’t just accuracy. It uses fewer tokens to complete the same tasks, making it both smarter and more efficient.
Early testers from Cursor, GitHub, and NVIDIA report that GPT-5.5 shows stronger “conceptual clarity,” understanding why something is failing and where the fix needs to land. An NVIDIA engineer said losing access to the model “feels like I’ve had a limb amputated.”
Knowledge Work and Computer Use#
GPT-5.5 scores 84.9% on GDPval (testing agents across 44 occupations) and 78.7% on OSWorld-Verified (operating real computer environments). OpenAI reports 85% of the company now uses Codex weekly across engineering, finance, comms, and marketing.
One standout example: Finance used Codex to review 24,771 K-1 tax forms (71,637 pages), accelerating the task by two weeks compared to the prior year.
Scientific Research Breakthroughs#
The model showed clear gains on GeneBench (multi-stage genetics data analysis) and BixBench (bioinformatics). Perhaps most impressive: an internal version helped discover a new proof about Ramsey numbers, later verified in Lean.
Efficiency Without Latency Trade-offs#
GPT-5.5 matches GPT-5.4’s per-token latency despite being more capable. It was co-designed for NVIDIA GB200 and GB300 NVL72 systems, and the model itself helped optimize the inference infrastructure that serves it.
Pricing and Availability#
Available now to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex. API pricing: $5 per 1M input tokens and $30 per 1M output tokens, with a 1M context window. GPT-5.5 Pro comes at $30/$180 for even harder tasks.
While priced higher than GPT-5.4, OpenAI notes it delivers better results with fewer tokens for most users.
The Bottom Line#
GPT-5.5 represents a meaningful step forward in agentic AI, particularly for coding, knowledge work, and scientific research. The combination of improved intelligence, better token efficiency, and maintained latency makes it a compelling upgrade. For enterprises already building on OpenAI’s platform, this model makes the case that AI agents are moving from demos to real production workflows.
Read the full announcement at openai.com.