GPT-5.6 Luna (Max)
OpenAIGPTProprietary
Description
GPT-5.6 Luna is the fastest and most cost-efficient tier of OpenAI's GPT-5.6 family, designed for cost-sensitive, high-volume workloads. It brings strong capability at OpenAI's lowest GPT-5.6 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
38
general
69
coding
91
reasoning
68
science
60
agents
85
multimodal
Classements
| Domaine | #Rang | Score | Source |
|---|---|---|---|
| Capacité agentique | 90 | 37.0 | LS |
| Classement codage | 28 | 90.0 | AA |
| Classement général | 80 | 63.0 | AA |
| Raisonnement mathématique | 37 | 87.0 | LB |
| Classement multimodal | 97 | 46.0 | LS |
| Raisonnement | 28 | 86.0 | LB |
| Science | 37 | 79.0 | AA |
Scores de benchmarks (LLM Stats)
(LLM Stats (zeroeval))Agents
Connectors
99.9%Aut.
Search and Function-Calling
89.7%Aut.
Capture-the-Flag Challenges (Internal)
85.2%Aut.
Toolathlon
53.4%Aut.
LifeSciBench
51.2%Aut.
Internal Research Debugging Evaluation
50.8%Aut.
Agents' Last Exam
50.3%Aut.
SEC-bench Pro
48.9%Aut.
RSI Index
41.9%Aut.
Big Finance Bench
36.0%Aut.
Management Consulting Tasks (Internal)
35.4%Aut.
PostTrainBench Lite
29.6%Aut.
KernelGen 1P
22.4%Aut.
AutomationBench
14.9%Aut.
ExploitGym
12.4%Aut.
GeneBench-Pro
10.8%Aut.
NanoGPT
1.7%Aut.
Code
BenchCAD (with Python tool)
73.9%Aut.
DeepSWE
67.2%Aut.
DeepSWE 1.1
67.0%
BenchCAD
63.1%Aut.
ExploitBench
33.2%Aut.
General
Artificial Analysis
51.0%
GDP.pdf
22.7%Aut.
Healthcare
HealthBench Consensus
95.1%Aut.
HealthBench
55.8%Aut.
HealthBench Professional
55.7%Aut.
HealthBench Hard
32.0%Aut.
Long Context
MRCR v2 (8-needle, 512K-1M)
41.3%Aut.
MRCR v2 (8-needle)
41.3%Aut.
Math
FrontierMath
78.6%Aut.
FrontierMath Tier 4 (v2)
58.5%Aut.
Multimodal
OSWorld 2.0
45.6%Aut.
Reasoning
GPQANYU + Cohere + Anthropic (2023)
92.3%Aut.
Terminal-Bench 2.1
84.7%Aut.
BrowseCompOpenAI (2025)
83.3%Aut.
Graphwalks BFS >128k
81.3%Aut.
SWE-Bench ProPrinceton NLP (2024)
62.7%Aut.
Graphwalks BFS 1M
51.2%Aut.
FrontierCode 1.1
39.8%
MedChemBench (Internal)
30.4%Aut.
Terminal-Bench 4.0
17.3%
ARC-AGI-3
0.2%Aut.
Vision
MMMU-Pro (with tools)
79.5%Aut.
MMMU-Pro
78.4%Aut.
Indices d'évaluation AA
(Artificial Analysis)Gpqa(NYU + Cohere + Anthropic (2023))91.1
Lcr(Artificial Analysis)83.7
Terminalbench V2 180.9
Coding Index(Artificial Analysis)71.4
Scicode(UIUC + Argonne National Lab (2024))53.6
Hle(Center for AI Safety + Scale AI (2025))39.5
Intelligence Index(Artificial Analysis)37.3
Tau Banking31.1
Terminalbench V4 011.6
Scores par catégorie LLM Stats
(LLM Stats (zeroeval))Physics90
Biology90
Chemistry90
Search80
Math70
Spatial Reasoning70
Multimodal60
Tool Calling60
Vision60
Long Context50
Reasoning50
General50
Healthcare50
Agents50
Code50
Safety40
Finance40
Science30
Systems20
Tarification
Prix d'entrée$0.2 / 1M tokens
Prix de sortie$1.2 / 1M tokens
Prix mixte (3:1)$0.45 / 1M tokens
Prix de lecture cache$0.02 / 1M tokens
Prix d'écriture cache$0.25 / 1M tokens
Vitesse
Tokens/sec0.0
Délai du premier token0.00s
Temps de réponse0.00s
Classement des Prix par Fournisseur
Classement des Prix par Fournisseur
2 fournisseurs
Moins cher: OpenAIPlus cher: Neon
FournisseurEntréeSortie
1OpenAIMoins cher
$0
$0
2Neon
$0.2
$1.2
Comparer les prix entre différents fournisseurs API pour ce modèle.