Saltar al contenido principal
AITier

GPT-5.5 (Xhigh)

OpenAIGPTProprietary

Descripción

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.

Fecha de lanzamiento
2026-04-23
Parámetros
—
Longitud del contexto
1.1M
Modalidades
image, pdf, text

Radar de capacidades

40
general
72
coding
93
reasoning
71
science
80
agents
85
multimodal

Rankings

Dominio#PosiciónPuntuaciónFuente
Capacidad agéntica38
50.0
LS
Ranking de codificación34
93.0
AA
Ranking general34
79.0
AA
Ranking multimodal21
62.0
LS
Ciencia35
84.0
AA

Puntuaciones de benchmarks (LLM Stats)

(LLM Stats (zeroeval))

Agents

BixBench80.5%Aut.
Finance Agent60.0%Aut.
Toolathlon55.6%Aut.
OfficeQA Pro54.1%Aut.
Finance Agent v251.8%
GeneBench25.0%Aut.

Code

CyberGym81.8%Aut.
FrontierSWE73.0%
DeepSWE 1.167.0%

Communication

Tau2 Telecom98.0%Aut.

General

GDPval-MM84.9%Aut.

Legal

Legal Agent Benchmark2.1%

Long Context

MRCR v2 (8-needle)74.0%Aut.

Math

LiveBench80.7%
FrontierMath35.4%Aut.

Multimodal

OSWorld-Verified78.7%Aut.

Reasoning

ARC-AGI95.0%Aut.
GPQANYU + Cohere + Anthropic (2023)93.6%Aut.
ARC-AGI v285.0%Aut.
BrowseCompOpenAI (2025)84.4%Aut.
Terminal-Bench 2.0Stanford × Laude Institute (2026)82.7%Aut.
MCP Atlas75.3%Aut.
SWE-Bench ProPrinceton NLP (2024)58.6%Aut.
Graphwalks parents >128k58.5%Aut.
Humanity's Last Exam52.2%Aut.
Graphwalks BFS >128k45.4%Aut.
FrontierCode 1.143.0%

Vision

MMMU-Pro83.2%Aut.

Índices de evaluación AA

(Artificial Analysis)
Tau2(Sierra + U Toronto + Vector Institute (2025))
93.9
Gpqa(NYU + Cohere + Anthropic (2023))
93.5
Lcr(Artificial Analysis)
84.3
Terminalbench V2 1
84.3
Ifbench(Google Research (2023))
75.9
Coding Index(Artificial Analysis)
74.9
Terminalbench Hard(Stanford × Laude Institute (2026))
60.6
Scicode(UIUC + Argonne National Lab (2024))
55.8
Hle(Center for AI Safety + Scale AI (2025))
45.8
Tau Banking
39.0
Intelligence Index(Artificial Analysis)
38.4
Terminalbench V4 0
14.6

Puntuaciones por categoría LLM Stats

(LLM Stats (zeroeval))
Communication
100
Physics
90
Biology
90
Chemistry
90
Multimodal
80
Safety
80
Search
80
Tool Calling
80
Vision
80
Reasoning
70
Spatial Reasoning
70
Finance
70
General
70
Code
70
Long Context
60
Math
60
Agents
60
Legal
0

Precios

Precio de entrada$5 / 1M tokens
Precio de salida$30 / 1M tokens
Precio mixto (3:1)$11.25 / 1M tokens
Precio de lectura caché$0.5 / 1M tokens

Velocidad

Tokens/seg0.0
Retraso del primer token0.00s
Tiempo hasta la respuesta0.00s

Ranking de Precios por Proveedor

Ranking de Precios por Proveedor

3 proveedores

Más barato: OpenAIMás caro: Neon
ProveedorEntradaSalida
1OpenAIMás barato
$0.00001
$0.00003
2FrogBot
$2.5
$15
3Neon
$5
$30

Comparar precios entre diferentes proveedores de API para este modelo.

Fuentes externas

LLM StatsArtificial Analysis