GPT-5.6 Luna
GPTCompact
ThinkingTool UseVisionStructured Output
About this model
GPT-5.6 Luna is a fast, cost-efficient model in OpenAI's GPT-5.6 series. It is suited for high-volume, latency-sensitive tasks such as chat, classification, and lightweight agentic workflows, providing capable reasoning for...
Performance Tier
Compact
GPT-5.6 Luna is a compact model from GPT : optimized for speed and affordability.
Small, fast, and affordable. Optimized for speed and low cost, great for high-volume or simple tasks.
Pricing
This model is included in Elosia plansAffordable
Low cost. Suitable for sustained use and high-volume interactions.
| Type | per 1M tokens |
|---|---|
| Input (prompt) | $0.100 |
| Output (completion) | $0.600 |
| Cache read | $0.010 |
| Cache write | $0.125 |
Capabilities
Context Length1.1M
Max Output Tokens128K
TokenizerGPT
Inputfile, image, text
Outputtext
Release DateJuly 9, 2026
Benchmarks
General Intelligence
MMLU
Not reported
Mathematics
MATH-500
Not reported
Programming
HumanEval
Not reported
Reasoning
ARC-AGI-2
29.3%
Agentic
SWE-bench Pro
62.7%
Terminal-Bench 2.1
84.7%
Where does GPT-5.6 Luna stand?
Compare its performance index against every other model.
Recommended Use Cases
General ChatCodingData ExtractionSummarizationCustomer Support
Strengths
- Fastest, most affordable tier of the GPT-5.6 family at $1 / $6 per M tokens
- 1M-token context window, uncommon at this price point, for large documents and long sessions
- Capable agentic coding for its class: 84.7 on Terminal-Bench 2.1 and 62.7 on SWE-bench Pro
- Multimodal (text, image, file) input with native tool calling, web search and function calling
Limitations
- Weakest abstract reasoning of the family (ARC-AGI-2 29.3 at high effort)
- Long-context recall is modest relative to its 1M window
- Only agentic benchmarks published by OpenAI; no MMLU, GPQA or math scores for comparison
Frequently asked questions
Resources
This model may use your data for training
