Models · OpenAI

GPT-5.5

Closed 2026-04-23

Lab: OpenAI
Released: 2026-04-23
Parameters: unknown
Context window: 1.05M
Input: text, image
Output: text
Weights: Closed
Official page: developers.openai.com ↗

Benchmark scores

ARC-AGI 2 reasoning 85%

85% 2026-04-25 OpenAI releases GPT-5.5, a ground-up retrained model with native omnimodal processing

BrowseComp agentic 84.4%

84.4% 2026-04-23 OpenAI releases GPT-5.5 with agentic capabilities and doubled API pricing

FrontierMath math 51.7%

51.7% 2026-04-25 OpenAI releases GPT-5.5, a ground-up retrained model with native omnimodal processing
35.4% 2026-04-23 OpenAI releases GPT-5.5 with agentic capabilities and doubled API pricing

Humanity's Last Exam reasoning 41.4%

41.4% 2026-04-25 OpenAI releases GPT-5.5, a ground-up retrained model with native omnimodal processing

OSWorld agentic 78.7%

78.7% 2026-04-25 OpenAI releases GPT-5.5, a ground-up retrained model with native omnimodal processing

SWE-bench Pro agentic 58.6%

58.6% 2026-04-23 OpenAI releases GPT-5.5 with agentic capabilities and doubled API pricing
58.6% 2026-04-23 OpenAI releases GPT-5.5 and GPT-5.5 Pro in API

Terminal-Bench agentic 82.7%

82.7% 2026-04-25 OpenAI releases GPT-5.5, a ground-up retrained model with native omnimodal processing
82.7% 2026-04-23 OpenAI releases GPT-5.5 and GPT-5.5 Pro in API
82.7% 2026-04-23 OpenAI releases GPT-5.5 with agentic capabilities and doubled API pricing

τ²-bench agentic 98%

98% 2026-04-25 OpenAI releases GPT-5.5, a ground-up retrained model with native omnimodal processing

Related articles