GPT-5.6 Luna vs GPT-5.6 Terra: Specs & Benchmark Comparison
GPT-5.6 Luna and GPT-5.6 Terra both come from OpenAI, and both were released in July 2026.
The two models share 45 published benchmarks. GPT-5.6 Terra leads on 41 of them, GPT-5.6 Luna on 3. The widest gaps are on MRCR v2 (8-needle), where GPT-5.6 Terra scores 89.6% against 41.3%; MRCR v2 (8-needle, 512K-1M), where GPT-5.6 Terra scores 72.5% against 41.3%. Averaged across everything we track, GPT-5.6 Luna sits at 51.5% and GPT-5.6 Terra at 59.0%.
GPT-5.6 Luna is the cheaper API at $0.20 per million input tokens and $1.20 per million output tokens, roughly 10 times cheaper than GPT-5.6 Terra at $2 and $12. Both accept a context window of about 1M tokens. Both accept text and images as input.
| Characteristic | GPT-5.6 Luna | GPT-5.6 Terra |
|---|---|---|
| Company | OpenAI | OpenAI |
| Release Date | July 9, 2026 | July 9, 2026 |
| Parameters | — | — |
| Multimodal | Yes | Yes |
| Context (input) | 1.1M | 1.1M |
| Context (output) | 128K | 128K |
| Input Price / 1M | $0.20 | $2.00 |
| Output Price / 1M | $1.20 | $12.00 |
| Average Score | 51.5% | 59.0% |
| Benchmarks | ||
| MRCR v2 (8-needle) | 41.3% | 89.6% |
| MRCR v2 (8-needle, 512K-1M) | 41.3% | 72.5% |
| KernelGen 1P | 22.4% | 49.2% |
| PostTrainBench Lite | 29.6% | 51.5% |
| Graphwalks BFS 1M | 51.2% | 71.2% |
| ExploitBench | 33.2% | 52.9% |
| Internal Research Debugging Evaluation | 50.8% | 67.8% |
| Big Finance Bench | 36.0% | 51.0% |
| RSI Index | 41.9% | 56.3% |
| NanoGPT | 1.7% | 14.5% |
| GeneBench-Pro | 10.8% | 23.3% |
| ExploitGym | 12.4% | 23.2% |
| FrontierMath Tier 4 (v2) | 58.5% | 68.3% |
| SEC-bench Pro | 48.9% | 57.7% |
| Capture-the-Flag Challenges (Internal) | 85.2% | 91.8% |
| FrontierMath | 78.6% | 84.9% |
| Search and Function-Calling | 89.7% | 94.6% |
| LifeSciBench | 51.2% | 56.0% |
| MedChemBench (Internal) | 30.4% | 35.0% |
| OSWorld 2.0 | 45.6% | 50.2% |
| Graphwalks BFS >128k | 81.3% | 76.9% |
| BenchCAD (with Python tool) | 73.9% | 78.2% |
| BrowseComp | 83.3% | 87.5% |
| Terminal-Bench 4.0 | 17.3% | 21.5% |
| Artificial Analysis | 51.0% | 55.0% |
| DeepSWE 1.1 | 67.0% | 70.0% |
| Terminal-Bench 2.1 | 84.7% | 87.4% |
| MMMU-Pro (with tools) | 79.5% | 82.0% |
| DeepSWE | 67.2% | 69.6% |
| MMMU-Pro | 78.4% | 80.7% |
| GDP.pdf | 22.7% | 24.7% |
| HealthBench Professional | 55.7% | 57.7% |
| Management Consulting Tasks (Internal) | 35.4% | 37.2% |
| FrontierCode 1.1 | 39.8% | 41.3% |
| HealthBench | 55.8% | 57.0% |
| BenchCAD | 63.1% | 62.3% |
| HealthBench Hard | 32.0% | 32.7% |
| SWE-Bench Pro | 62.7% | 63.4% |
| ARC-AGI-3 | 0.2% | 0.8% |
| GPQA | 92.3% | 92.9% |
| AutomationBench | 14.9% | 15.2% |
| Toolathlon | 53.4% | 53.1% |
| Agents' Last Exam | 50.3% | 50.4% |
| Connectors | 99.9% | 100.0% |
| HealthBench Consensus | 95.1% | 95.1% |
Visual Benchmark Comparison
Verdict
GPT-5.6 Luna leads in 1 out of 4 comparison categories.
Both models show comparable average scores: GPT-5.6 Luna — 0.5, GPT-5.6 Terra — 0.6.
GPT-5.6 Luna is 10.0x cheaper: input $0.20/1M vs $2.00/1M tokens.
Same context size: 1M tokens.
Both models were released around the same time: 7/9/2026 and 7/9/2026.
More About These Models
Frequently Asked Questions
Which is better for coding — GPT-5.6 Luna or GPT-5.6 Terra?
Which model is cheaper — GPT-5.6 Luna or GPT-5.6 Terra?
Which has a larger context window — GPT-5.6 Luna or GPT-5.6 Terra?
The GPT-5.6 Luna and GPT-5.6 Terra comparison is updated for 2026. Data includes benchmark results, API pricing, context window size and other specifications. For more detailed information, visit the GPT-5.6 Luna or GPT-5.6 Terra page. See also the complete list of AI model comparisons.