GPT-5.5

GPT-5.5, released on 23 April 2026, offers a one-million-token context window in the API, with output capped at 128,000 tokens. Input is text and images and output is text; audio and video remain outside the base API. In Codex the model runs with a 400,000-token window across the Plus, Pro, Business, Enterprise, Edu and Go plans. Its strength is long-horizon agentic work. OpenAI reports 82.7% on Terminal-Bench 2.0, 73.1% on its internal Expert-SWE long-horizon benchmark and 84.9% on GDPval; on FrontierMath Tier 4 the base model reaches 35.4% and the Pro tier 39.6%. Against GPT-5.4, released just seven weeks earlier, it improves on nine of ten shared benchmarks, with the largest gains on ARC-AGI-2 (+11.7 points), MCP Atlas (+8.1) and Terminal-Bench 2.0 (+7.6) — and it uses fewer tokens to finish a coding task while holding roughly the same per-token latency in real-world serving. The price moved sharply: five dollars per million input tokens and thirty per million output, double the GPT-5.4 rates, with GPT-5.5 Pro at thirty and one hundred eighty. Batch and Flex run at half the standard rate, Priority processing at 2.5×, and Codex offers a Fast mode that generates tokens 1.5× faster for 2.5× the cost. The later GPT-5.6 has its own entry in this catalogue; GPT-5.4 is not documented here yet.

▸Videos

No videos yet.

▸Reviews