GPT-5.5 (low)
OpenAIGPTProprietary
Описание
GPT-5.5 is OpenAI's smartest model yet, designed for real work across agentic coding, computer use, knowledge work, and early scientific research. It matches GPT-5.4 per-token latency in real-world serving while reaching a much higher level of intelligence and using significantly fewer tokens to complete the same tasks. GPT-5.5 supports a 1M-token context window in the API and a 400K-token context window in Codex, with state-of-the-art results on Terminal-Bench 2.0, OSWorld-Verified, GDPval, FrontierMath, and CyberGym.
Дата выхода
2026-04-23
Параметры
—
Длина контекста
1.1M
Модальности
image, pdf, text
Радар способностей
31
general
61
coding
91
reasoning
73
science
80
agents
85
multimodal
Рейтинги
| Домен | #Место | Оценка | Источник |
|---|---|---|---|
| Агентные возможности | 39 | 50.0 | LS |
| Рейтинг кодинга | 113 | 80.0 | AA |
| Общий рейтинг | 114 | 65.0 | AA |
| Мультимодальный рейтинг | 22 | 62.0 | LS |
| Наука | 102 | 73.0 | AA |
Оценки бенчмарков (LLM Stats)
(LLM Stats (zeroeval))Agents
BixBench
80.5%Сам.
Finance Agent
60.0%Сам.
Toolathlon
55.6%Сам.
OfficeQA Pro
54.1%Сам.
Finance Agent v2
51.8%
GeneBench
25.0%Сам.
Code
CyberGym
81.8%Сам.
FrontierSWE
73.0%
DeepSWE 1.1
67.0%
Communication
Tau2 Telecom
98.0%Сам.
General
GDPval-MM
84.9%Сам.
Legal
Legal Agent Benchmark
2.1%
Long Context
MRCR v2 (8-needle)
74.0%Сам.
Math
LiveBench
80.7%
FrontierMath
35.4%Сам.
Multimodal
OSWorld-Verified
78.7%Сам.
Reasoning
ARC-AGI
95.0%Сам.
GPQANYU + Cohere + Anthropic (2023)
93.6%Сам.
ARC-AGI v2
85.0%Сам.
BrowseCompOpenAI (2025)
84.4%Сам.
Terminal-Bench 2.0Stanford × Laude Institute (2026)
82.7%Сам.
MCP Atlas
75.3%Сам.
SWE-Bench ProPrinceton NLP (2024)
58.6%Сам.
Graphwalks parents >128k
58.5%Сам.
Humanity's Last Exam
52.2%Сам.
Graphwalks BFS >128k
45.4%Сам.
FrontierCode 1.1
43.0%
Vision
MMMU-Pro
83.2%Сам.
Индексы оценки AA
(Artificial Analysis)Gpqa(NYU + Cohere + Anthropic (2023))91.0
Tau2(Sierra + U Toronto + Vector Institute (2025))83.9
Lcr(Artificial Analysis)81.0
Terminalbench V2 165.5
Ifbench(Google Research (2023))64.4
Coding Index(Artificial Analysis)60.9
Terminalbench Hard(Stanford × Laude Institute (2026))52.3
Hle(Center for AI Safety + Scale AI (2025))32.7
Intelligence Index(Artificial Analysis)30.7
Tau Banking24.9
Оценки категорий LLM Stats
(LLM Stats (zeroeval))Communication100
Physics90
Biology90
Chemistry90
Multimodal80
Safety80
Search80
Tool Calling80
Vision80
Reasoning70
Spatial Reasoning70
Finance70
General70
Code70
Long Context60
Math60
Agents60
Legal0
Цены
Цена ввода$5 / 1M токенов
Цена вывода$30 / 1M токенов
Смешанная цена (3:1)$11.25 / 1M токенов
Цена чтения кэша$0.5 / 1M токенов
Скорость
Токенов/сек0.0
Задержка первого токена0.00s
Время до первого ответа0.00s
Рейтинг цен провайдеров
Рейтинг цен провайдеров
1 провайдеров
ПровайдерВводВывод
1OpenAI
$0.00001
$0.00003
Сравнение цен разных API-провайдеров для этой модели.