OpenAImodel systemcard

GPT-5.5

#6of 27 · general

incl. pre-release runs; Pro is a different product and excluded

1.1Mcontext

Proprietarynon-commercial

2026-04-23released

2025-12-01knowledge cutoff

Standing

dot = this model · shaded = the field's middle 60% · track = full field

General

67.2#6/27

Math

74.2#5/27

Coding

62.7#8/27

Health

40.2#15/26

Cyber

84.6#3/14

Analyst

90.0#2/11

Legal

30.9#20/26

Instruct

67.4#5/27

Agents

48.9#13/27

Professional

54.9#10/27

Multimodal

64.2#4/19

Hallucination

59.6#2/27

Value

55.7#9/27

Measured

≈ imputed · amber dot = lab-self-reported among the sources

GPQA Diamond93.6%#4

HLE52.2%#3

MMLU-Pro88.1%#10

LiveBench80.4%#3

SWE-bench V78.4%#14

LiveCodeBench85.3%#15

SciCode56.1%#6

CritPt27.1%#5

Terminal-Bench78.0%#8

APEX Agents38.5%#7

Arena1482#8

Vals Index68.0#8

Epoch Capabilities Index159.2#4

Operations

$0.43$ per solved task

1482arena elo · 47,572 votes

Series

OpenAI's models on the board, by release

GPT-5.52026-04-23 · 67GPT-5.6 Sol2026-07-09 · 85GPT-5.6 Terra2026-07-09 · 59GPT-5.6 Luna2026-07-09 · 50

Neighbors

nearest capability fingerprints · deltas describe the neighbor

GPT-5.6 Sol→ stronger coding GLM-5.2→ looser instructions Gemini 3.5 Flash→ weaker coding Kimi K3→ weaker math Qwen3.8 Max→ weaker math · weaker multimodal