Chat with GPT-5.6 Luna Now
GPT-5.6 Luna: High Performance at an Affordable Scale
GPT-5.6 Luna is the most cost-efficient model in OpenAI’s GPT-5.6 family, designed to deliver accessible intelligence for everyday work. It is engineered to enable you to run complex, agentic workflows with fewer tokens.
Luna outperforms Claude Opus 4.8 at roughly one-quarter the estimated cost. It nearly matches the peak performance of GPT-5.5, making it an ideal choice for scaling high-volume code generation and technical operations.
Key Technical Specs of GPT-5.6 Luna
- Context Window: Up to 1,050,000 tokens
- Max Output: 128,000 tokens
- Knowledge Cutoff: Feb 16, 2026
- API Input Pricing: $1.00 / 1M tokens
- API Output Pricing: $6.00 / 1M tokens
- Cached Input: $0.1 / 1M tokens
Main Features of GPT-5.6 Luna
Cost-Efficiency and Token Optimization
GPT-5.6 Luna delivers maximum value per dollar. Compared with Claude Opus 4.8, Luna requires half the output tokens and only one-third of the time to complete complex tasks, making it effective for production-level workloads.
High-Capability Coding
Despite being the most economical model in the GPT-5.6 family, Luna punches well above its weight class in software engineering. It excels at processing complex command-line workflows, allowing developers to build robust coding agents.
Desktop-Grade Knowledge Work
GPT-5.6 Luna brings agentic capabilities to everyday productivity tools. It efficiently navigates multi-step workflows, extracts vital insights from complex documents, and converts messy corporate data into professional, ready-to-use artifacts.
Comparison of GPT-5.6 Series
| Evaluation | GPT-5.6 Luna | GPT-5.6 Sol | GPT-5.6 Terra |
| Definition | Lightweight, cost-efficient variant | Highest reasoning ceiling | Balanced, all-round choice |
| Agents' Last Exam | 50.3% | 52.7% | 50.4% |
| Management Consulting Tasks | 35.4% | 43.2% | 37.2% |
| SWE-Bench Pro | 62.7% | 64.6% | 63.4% |
| Terminal Bench 2.1 | 84.7% | 88.8% | 87.4% |
| OSWorld 2.0 | 45.6% | 62.6% | 50.2% |
| GPQA Diamond | 92.3% | 94.6% | 92.9% |
Explore Our Other Models
Discover other state-of-the-art models you can chat with on HIX AI.