GPT-5.5
FrontierOpenAI•Released on 2026-04-23
OpenAI's latest frontier model released April 23, 2026. GPT-5.5 is the first fully retrained base model since GPT-4.5, built with a natively omnimodal architecture. Leads agentic workflows with 82.7% on Terminal-Bench 2.0 (state-of-the-art) and 84.9% GDPval, narrowly edging Anthropic's gated Claude Mythos Preview on Terminal-Bench.
Voice of the community
sample 45“GPT-5.5 narrowly beats Anthropic's Claude Mythos Preview on Terminal-Bench 2.0 — OpenAI's first fully retrained base model since GPT-4.5, built omnimodal from the ground up.”
“GPT-5.5 is a clean step up on every agentic benchmark — Terminal-Bench improving 7.6pp and OSWorld climbing 3.7pp, while matching the previous version's latency and using fewer tokens per task.”
“The three frontier models have different strengths: GPT-5.5 leads agentic workflows, Gemini 3.1 Pro leads reasoning, Claude Mythos leads coding.”
Core Specs
Scenario Scores
Pros & Cons
Pros
- +SOTA on Terminal-Bench 2.0 (82.7%)
- +Leads agentic benchmarks (GDPval 84.9%, OSWorld 78.7%)
- +922K context window
- +First fully retrained base since GPT-4.5
- +Matches GPT-5.4 latency with fewer tokens per task
Cons
- −$30 output price 2x of GPT-5.4 ($15)
- −Still trails Claude Mythos Preview on SWE-Bench (93.9% vs 82.7%)
- −Pro tier pricing ($30/$180) steep vs competitors
- −Community data limited (released 2 days ago)