GPT-5.6 Luna
Active Parameters
-
Context Length
1.05M
Modality
Multimodal
Architecture
Mixture of Experts (MoE)
License
-
Release Date
9 Jul 2026
Knowledge Cutoff
Feb 2026
Evaluation Benchmarks
Rank
#50
| Benchmark | Score | Rank |
|---|---|---|
Coding max | 0.83 | ⭐ 4 |
Data Analysis max | 0.78 | 19 |
Reasoning max | 0.86 | 21 |
Web Development high | 1518 | 25 |
Web Development max | 1518 | 25 |
Web Development Standard | 1518 | 25 |
Agent Arena high | 0.04 | 25 |
Agent Arena Standard | 0.02 | 35 |
General max | 0.74 | 25 |
Agentic Coding max | 0.48 | 26 |
LiveBench Average max | 0.74 | 27 |
Mathematics max | 0.87 | 36 |
General Text high | 1452 | 56 |
General Text Standard | 1450 | 58 |
Rankings
Overall Rank
#50
Coding Rank
#38
About GPT-5.6 Luna
OpenAI's GPT-5.6 Luna is a lightweight, ultrafast, and cost-effective tier designed for high-volume inference, subagent execution, and low-latency workflows.
Technical Specifications
Detailed architecture specifications are limited for proprietary models.
Resources
About GPT-5.6
OpenAI's GPT-5.6 series represents the next-generation frontier intelligence family structured into three tiered variants: Sol (flagship reasoning and autonomous agents), Terra (balanced everyday production workhorse), and Luna (lightweight, ultrafast, cost-effective inference). It introduces new max and ultra (sub-agent coordination) reasoning modes, native programmatic tool calling, a 1M token context window with 128k output tokens, and a February 16, 2026 knowledge cutoff.