GPT-5.6 LunavsGPT-5.5
Across 10 shared benchmarks, GPT-5.5 leads overall: GPT-5.6 Luna wins 4, GPT-5.5 wins 6, with 0 ties and an average score difference of -3.60.
GPT-5.6 Luna
OpenAI · 2026-06-26 · Reasoning model
GPT-5.5
OpenAI · 2026-04-23 · Reasoning model
GPT-5.6 Luna4 wins(40%)(60%)6 winsGPT-5.5
Benchmark scores
Grouped by capability, sorted by largest gap within each. 10 shared benchmarks.
Coding and Software Engineer
GPT-5.6 Luna 2/3| Benchmark | GPT-5.6 Luna | GPT-5.5 | Diff |
|---|---|---|---|
| Text Arena (Coding) | 1,52315 / 35极高强度思考（无工具） | 1,50518 / 35极高强度思考（无工具） | +18.20 |
| WeirdML v2 | 60.8627 / 52Thinking High (With Tools) | 67.1522 / 52Normal (With Tools) | -6.29 |
| DeepSWE | 67.206 / 27极高强度思考（工具） | 677 / 27极高强度思考（工具） | +0.20 |
General Knowledge
GPT-5.5 2/3| Benchmark | GPT-5.6 Luna | GPT-5.5 | Diff |
|---|---|---|---|
| ARC-AGI-2 | 59.5020 / 62最高（无工具） | 852 / 62Thinking High (No Tools) | -25.50 |
| ARC-AGI | 8819 / 68最高（无工具） | 955 / 68极高强度思考（无工具） | -7 |
| ARC-AGI-3 | 0.203 / 9最高（无工具） | 05 / 9Thinking High (No Tools) | +0.20 |
Math and Reasoning
GPT-5.5 2/2| Benchmark | GPT-5.6 Luna | GPT-5.5 | Diff |
|---|---|---|---|
| FrontierMath Tier 4 v2 | 60.987 / 34最高（无工具） | 72.505 / 34极高强度思考（无工具） | -11.52 |
| FrontierMath v2 | 82.118 / 34最高（无工具） | 85.266 / 34极高强度思考（无工具） | -3.16 |
AI Agent - Tool Usage
GPT-5.6 Luna 1/1| Benchmark | GPT-5.6 Luna | GPT-5.5 | Diff |
|---|---|---|---|
| Terminal-Bench 2.1 | 84.7010 / 44最高（无工具） | 83.4013 / 44Thinking High (With Tools) | +1.30 |
General Evaluation
GPT-5.5 1/1| Benchmark | GPT-5.6 Luna | GPT-5.5 | Diff |
|---|---|---|---|
| GPQA Diamond | 91.6027 / 226最高（无工具） | 947 / 226极高强度思考（无工具） | -2.40 |
Specs
| Field | GPT-5.6 Luna | GPT-5.5 |
|---|---|---|
| Publisher | OpenAI | OpenAI |
| Release date | 2026-06-26 | 2026-04-23 |
| Model type | Reasoning model | Reasoning model |
| Architecture | Dense | Dense |
| Parameters | Not available | Not available |
| Context length | 1.05M | 1000K |
| Max output | 128K | 128K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | GPT-5.6 Luna | GPT-5.5 |
|---|---|---|
| Text input | $1 / 1M tokens | $0.5 / 1M tokens |
| Text output | $6 / 1M tokens | $30 / 1M tokens |
| Cache read | $0.1 / 1M tokens | $0.5 / 1M tokens |
| Cache write | $1.25 / 1M tokens | $6.25 / 1M tokens |
Summary
- GPT-5.6 Lunaleads in:Coding and Software Engineer (2/3), AI Agent - Tool Usage (1/1)
- GPT-5.5leads in:General Knowledge (2/3), Math and Reasoning (2/2), General Evaluation (1/1)
On average across the 10 shared benchmarks, GPT-5.5 scores 3.60 higher.
Largest single-benchmark gap: ARC-AGI-2 — GPT-5.6 Luna 59.50 vs GPT-5.5 85 (-25.50).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.