GPT-5.6 LunavsGPT-5.4
Across 9 shared benchmarks, GPT-5.6 Luna leads overall: GPT-5.6 Luna wins 6, GPT-5.4 wins 3, with 0 ties and an average score difference of +8.39.
GPT-5.6 Luna
OpenAI · 2026-06-26 · Reasoning model
GPT-5.4
OpenAI · 2026-03-05 · Multimodal model
GPT-5.6 Luna6 wins(67%)(33%)3 winsGPT-5.4
Benchmark scores
Grouped by capability, sorted by largest gap within each. 9 shared benchmarks.
Coding and Software Engineer
GPT-5.6 Luna 3/3| Benchmark | GPT-5.6 Luna | GPT-5.4 | Diff |
|---|---|---|---|
| Text Arena (Coding) | 1,52315 / 35极高强度思考（无工具） | 1,45724 / 35Thinking High (No Tools) | +65.71 |
| DeepSWE | 67.206 / 27极高强度思考（工具） | 5219 / 27极高强度思考（工具） | +15.20 |
| WeirdML v2 | 60.8627 / 52Thinking High (With Tools) | 57.4431 / 52Normal (With Tools) | +3.42 |
General Knowledge
GPT-5.4 2/3| Benchmark | GPT-5.6 Luna | GPT-5.4 | Diff |
|---|---|---|---|
| ARC-AGI-2 | 59.5020 / 62最高（无工具） | 77.109 / 62Normal (No Tools) | -17.60 |
| ARC-AGI | 8819 / 68最高（无工具） | 93.709 / 68Normal (No Tools) | -5.70 |
| ARC-AGI-3 | 0.203 / 9最高（无工具） | 07 / 9Thinking High (No Tools) | +0.20 |
Math and Reasoning
GPT-5.6 Luna 2/2| Benchmark | GPT-5.6 Luna | GPT-5.4 | Diff |
|---|---|---|---|
| FrontierMath Tier 4 v2 | 60.987 / 34最高（无工具） | 4910 / 34极高强度思考（无工具） | +11.98 |
| FrontierMath v2 | 82.118 / 34最高（无工具） | 78.6010 / 34极高强度思考（无工具） | +3.51 |
General Evaluation
GPT-5.4 1/1| Benchmark | GPT-5.6 Luna | GPT-5.4 | Diff |
|---|---|---|---|
| GPQA Diamond | 91.6027 / 226最高（无工具） | 92.8020 / 226极高强度思考（无工具） | -1.20 |
Specs
| Field | GPT-5.6 Luna | GPT-5.4 |
|---|---|---|
| Publisher | OpenAI | OpenAI |
| Release date | 2026-06-26 | 2026-03-05 |
| Model type | Reasoning model | Multimodal model |
| Architecture | Dense | Dense |
| Parameters | Not available | Not available |
| Context length | 1.05M | 1M |
| Max output | 128K | 125K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | GPT-5.6 Luna | GPT-5.4 |
|---|---|---|
| Text input | $1 / 1M tokens | $2.5 / 1M tokens |
| Text output | $6 / 1M tokens | $15 / 1M tokens |
| Cache read | $0.1 / 1M tokens | Not public |
| Cache write | $1.25 / 1M tokens | $0.25 / 1M tokens |
Summary
- GPT-5.6 Lunaleads in:Coding and Software Engineer (3/3), Math and Reasoning (2/2)
- GPT-5.4leads in:General Knowledge (2/3), General Evaluation (1/1)
On average across the 9 shared benchmarks, GPT-5.6 Luna scores 8.39 higher.
Largest single-benchmark gap: Text Arena (Coding) — GPT-5.6 Luna 1,523 vs GPT-5.4 1,457 (+65.71).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.