GPT-5.6 Luna
GPT-5.6 Luna is OpenAI’s high-speed, cost-optimized model tier within the GPT-5.6 family (alongside Sol and Terra). Serving as the lightweight nano-tier engine of the suite, Luna is engineered for high-throughput, latency-sensitive applications, high-volume classification, and streaming sub-agent orchestration.
Model Overview
Luna bridges the gap between ultra-low unit economics and frontier-class context processing. It enables developers to run continuous, large-scale agentic pipelines and full-document passes without incurring the cost overhead of flagship reasoning engines.
- Developer: OpenAI
- Model Tier: Luna (Fast / High-Throughput Efficiency)
- Release Date: July 9, 2026
- Primary Target Use Cases: Real-time interactive assistants, high-volume classification, structured field extraction, sub-agent execution loops, and long-document processing.
Key Capabilities & Performance Profile
Ultra-Low Unit Economics
- High-Volume Viability: Designed specifically for features that touch every request, user turn, or document pass, keeping operational costs low across high-traffic production APIs.
- Token Efficiency: Inherits the GPT-5.6 family’s token-reduction optimizations, using fewer generation tokens to reach conclusions on intermediate reasoning steps.
Extended Context & Long Generation
- 1.1 Million Token Window: Matches the 1,100,000-token context window of its larger siblings (Terra and Sol), allowing full codebases, logs, and lengthy spec sheets to be processed in a single pass.
- High Output Headroom: Supports output lengths up to 272,000 tokens per request for extended code outputs or document generation.
Sub-Agent Orchestration & Native Tools
- Agent Pipeline Node: Functions as a low-cost worker model in multi-agent architectures, executing routine classification, tool formatting, and initial task routing before escalating complex edge cases to GPT-5.6 Sol.
- Responses API Support: Native support for web search, file search, code execution, computer use, and Model Context Protocol (MCP) integrations.
Technical Specifications
| Parameter |
Specification |
| Model ID |
gpt-5.6-luna / openai/gpt-5.6-luna |
| Context Window |
1,100,000 tokens (1.1M) |
| Max Output Tokens |
272,000 tokens |
| Input Modalities |
Text, Code, Images (Vision) |
| Output Modalities |
Text, Code, Structured JSON |
| Native Features |
Function Calling, Prompt Caching, Reasoning Tokens, Computer Use, MCP Tools |
GPT-5.6 Family Tiering
| Tier |
Role & Capability Target |
| GPT-5.6 Sol |
Flagship: Maximum reasoning, top-tier coding, defensive cybersecurity, and deep multi-step architecture. |
| GPT-5.6 Terra |
Balanced Workhorse: Mid-tier option balancing intelligence and cost for standard enterprise production traffic. |
| GPT-5.6 Luna |
Ultra-Fast Efficiency: High-throughput, low-latency engine designed for cost-sensitive, high-volume workloads. |