GPT-5.6 Terra (Max)
OpenAIGPTProprietary
Description
GPT-5.6 Terra is the balanced tier of OpenAI's GPT-5.6 family, designed for workloads that balance intelligence and cost. It delivers performance competitive with GPT-5.5 at roughly half the price, supports max reasoning effort, and has a 1.05M-token context window.
Date de sortie
2026-07-09
Paramètres
—
Longueur du contexte
1.1M
Modalités
image, pdf, text
Radar de capacités
42
general
73
coding
93
reasoning
69
science
60
agents
85
multimodal
Classements
| Domaine | #Rang | Score | Source |
|---|---|---|---|
| Capacité agentique | 71 | 44.0 | LS |
| Classement codage | 22 | 92.0 | AA |
| Classement général | 13 | 79.0 | AA |
| Raisonnement mathématique | 13 | 95.0 | LB |
| Classement multimodal | 77 | 52.0 | LS |
| Raisonnement | 8 | 91.0 | LB |
| Science | 24 | 82.0 | AA |
Scores de benchmarks (LLM Stats)
(LLM Stats (zeroeval))Agents
Connectors
100.0%Aut.
Search and Function-Calling
94.6%Aut.
Capture-the-Flag Challenges (Internal)
91.8%Aut.
Internal Research Debugging Evaluation
67.8%Aut.
SEC-bench Pro
57.7%Aut.
RSI Index
56.3%Aut.
LifeSciBench
56.0%Aut.
Toolathlon
53.1%Aut.
PostTrainBench Lite
51.5%Aut.
Big Finance Bench
51.0%Aut.
Agents' Last Exam
50.4%Aut.
KernelGen 1P
49.2%Aut.
Management Consulting Tasks (Internal)
37.2%Aut.
GeneBench-Pro
23.3%Aut.
ExploitGym
23.2%Aut.
AutomationBench
15.2%Aut.
NanoGPT
14.5%Aut.
Code
BenchCAD (with Python tool)
78.2%Aut.
DeepSWE 1.1
70.0%
DeepSWE
69.6%Aut.
BenchCAD
62.3%Aut.
ExploitBench
52.9%Aut.
General
Artificial Analysis
55.0%
GDP.pdf
24.7%Aut.
Healthcare
HealthBench Consensus
95.1%Aut.
HealthBench Professional
57.7%Aut.
HealthBench
57.0%Aut.
HealthBench Hard
32.7%Aut.
Long Context
MRCR v2 (8-needle)
89.6%Aut.
MRCR v2 (8-needle, 512K-1M)
72.5%Aut.
Math
FrontierMath
84.9%Aut.
FrontierMath Tier 4 (v2)
68.3%Aut.
Multimodal
OSWorld 2.0
50.2%Aut.
Reasoning
GPQANYU + Cohere + Anthropic (2023)
92.9%Aut.
BrowseCompOpenAI (2025)
87.5%Aut.
Terminal-Bench 2.1
87.4%Aut.
Graphwalks BFS >128k
76.9%Aut.
Graphwalks BFS 1M
71.2%Aut.
SWE-Bench ProPrinceton NLP (2024)
63.4%Aut.
FrontierCode 1.1
41.3%
MedChemBench (Internal)
35.0%Aut.
Terminal-Bench 4.0
21.5%
ARC-AGI-3
0.8%Aut.
Vision
MMMU-Pro (with tools)
82.0%Aut.
MMMU-Pro
80.7%Aut.
Indices d'évaluation AA
(Artificial Analysis)Gpqa(NYU + Cohere + Anthropic (2023))92.5
Terminalbench V2 188.0
Tau2(Sierra + U Toronto + Vector Institute (2025))86.3
Lcr(Artificial Analysis)83.0
Coding Index(Artificial Analysis)76.7
Ifbench(Google Research (2023))71.2
Terminalbench Hard(Stanford × Laude Institute (2026))57.6
Scicode(UIUC + Argonne National Lab (2024))55.0
Hle(Center for AI Safety + Scale AI (2025))42.9
Intelligence Index(Artificial Analysis)42.1
Tau Banking40.2
Terminalbench V4 035.4
Scores par catégorie LLM Stats
(LLM Stats (zeroeval))Physics90
Search90
Biology90
Chemistry90
Long Context80
Math80
Spatial Reasoning70
Multimodal60
Reasoning60
Safety60
General60
Healthcare60
Agents60
Code60
Tool Calling60
Vision60
Science40
Finance40
Systems40
Tarification
Prix d'entrée$2 / 1M tokens
Prix de sortie$12 / 1M tokens
Prix mixte (3:1)$4.5 / 1M tokens
Prix de lecture cache$0.2 / 1M tokens
Prix d'écriture cache$2.5 / 1M tokens
Vitesse
Tokens/sec102.2
Délai du premier token97.36s
Temps de réponse97.36s
Classement des Prix par Fournisseur
Classement des Prix par Fournisseur
2 fournisseurs
Moins cher: OpenAIPlus cher: Neon
FournisseurEntréeSortie
1OpenAIMoins cher
$0
$0.00001
2Neon
$2
$12
Comparer les prix entre différents fournisseurs API pour ce modèle.