GPT-5.6 Luna
GPT-5.6 Luna is an AI model from OpenAI, released on 2026-06-26.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
Model basics
Open source & experience
Official resources
API details
| Type | Condition | Input | Output |
|---|---|---|---|
| Text | - | $0.200/ 1M | $1.20/ 1M |
| Type | Condition | Input | Output |
|---|---|---|---|
| Text | - | $0.400/ 1M | $1.80/ 1M |
| Type | TTL | Write | Read |
|---|---|---|---|
| Text | 30m | $0.250/ 1M | $0.020/ 1M |
“—” means the modality is not billed in that direction, or the vendor has not published a price for it.
Benchmark Results
GPT-5.6 Luna currently shows benchmark results led by GPQA Diamond (34 / 274, score 91.60), FrontierMath v2 (8 / 58, score 82.11), Creative Writing (17 / 106, score 1825.80). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.
General Knowledge
4 evaluationsGeneral Evaluation
3 evaluationsWriting and Creative Capabilities
1 evaluationsAI Agent - Tool Usage
4 evaluationsCoding and Software Engineer
4 evaluationsMath and Reasoning
4 evaluationsCompare with other models
Publisher
Model Overview
GPT-5.6 Luna
GPT-5.6 Luna is OpenAI's most cost-efficient tier in the GPT-5.6 family, first previewed on June 26, 2026 and made generally available on July 9, 2026. It's built for high-throughput, cost-sensitive tasks — summarization, drafting, routine automation — and is the fastest and cheapest of the three GPT-5.6 models. The family also includes Sol (the flagship) and Terra (the balanced mid-tier).
On Artificial Analysis's Intelligence Index (max reasoning), Luna scores 51 at roughly one-fifth Sol's cost per task (~$0.21 vs ~$1.04), delivering about 24 benchmark points per estimated API dollar. It reaches 84.70 on Terminal-Bench 2.1, 67.20 on DeepSWE, 74.6 on the AA Coding Agent Index, and 50.3 on Agents' Last Exam. Long-context recall is comparatively weak (41.3% on MRCR), so it's best reserved for shorter, well-scoped jobs rather than large-document workflows.
Luna shares OpenAI's GPT-5.6 system card with Sol and Terra, including "High" Preparedness Framework ratings for cybersecurity and biological/chemical risk; see the GPT-5.6 Sol page for the detailed safety-evaluation breakdown.
DataLearner on WeChat
Follow DataLearner on WeChat for AI model updates and research notes.
