GPT-5.6 Sol vs GPT-6 Luna
GPT-5.6 Sol Verdict
- GPT-5.6 Sol (max) leads on quality: 68.4 vs 61.2.
- GPT-5.6 Sol (max) is stronger in agents & tools, coding, composite, instruction following, knowledge, long context, maths, multimodal.
- GPT-6 Luna (max) is stronger in reasoning.
- GPT-6 Luna (max) is 40× cheaper ($0.200 vs $8.00 per 1M blended).
- GPT-6 Luna (max) streams 1.8× faster (157 vs 85 tokens per second).
| Metric | GPT-5.6 Sol (max) | GPT-6 Luna (max) |
|---|---|---|
| BenchLeader Index | 68.4 | 61.2 |
| Agents & tools score | 66.6 | – |
| Coding score | 67.0 | – |
| Composite score | 77.5 | 73.9 |
| Instruction following score | 71.0 | – |
| Knowledge score | 66.2 | 63.2 |
| Long context score | 67.9 | 67.5 |
| Maths score | 74.3 | – |
| Multimodal score | 67.3 | 59.2 |
| Reasoning score | 73.7 | 76.1 |
| Blended price $/M | $8.00 | $0.200 |
| Output speed | 85 tok/s | 157 tok/s |
| Time to first answer | 124.3 s | 142.6 s |
| Context window | 1.1M | 1.1M |
| GPQA Diamond | 93.5% | – |
| FrontierMath Tiers 1–3 | 89.1% | – |
| FrontierMath Tier 4 | 82.9% | – |
| OTIS Mock AIME | 100.0% | – |
| SimpleQA Verified | 69.7% | – |
| Terminal-Bench | 37.3% | – |
| OSWorld-Verified 2.0 | 27.3% | – |
| SciCode | 57.1% | – |
| WeirdML | 87.0% | – |
| ProofBench | 83.0% | – |
| LiveBench | 81.0% | – |
| LiveBench Reasoning | 91.7% | – |
| LiveBench Coding | 83.9% | – |
| LiveBench Agentic Coding | 56.2% | – |
| LiveBench Mathematics | 96.2% | – |
| LiveBench Data Analysis | 79.8% | – |
| LiveBench Language | 87.7% | – |
| LiveBench Instruction Following | 71.8% | – |
| AA Intelligence Index | 47.0 | 37.3 |
| IFBench | 72.7% | – |
| AA-LCR | 84.0% | 83.3% |
| MMMU-Pro | 83.4% | 75.5% |
| AA-Omniscience | 22.0 | 0.7 |
| Terminal-Bench Hard | 65.9% | – |
| GPQA Diamond (AA) | 94.1% | – |
| Humanity's Last Exam (AA) | 49.5% | 38.5% |
| SciCode (AA) | 57.1% | 54.6% |
| τ²-Bench Telecom (AA) | 85.1% | – |
| LiveCodeBench | 82.6% | – |
| MMLU-Pro | 89.1% | – |
| IOI | 91.2% | – |
| LegalBench | 87.0% | – |
| CorpFin | 64.4% | – |
| TaxEval | 74.8% | – |
| Terminal-Bench 2.1 (Vals) | 85.8% | – |
| SWE-bench (Vals) | 96.2% | – |
| GPQA Diamond (Vals) | 95.2% | – |
| Vals Index | 63.7 | – |
| PRBench Finance | 50.5% | – |
| PRBench Legal | 50.5% | – |
| ARC-AGI-1 | 96.5% | – |
| ARC-AGI-2 | 92.5% | – |
| ARC-AGI-3 | 7.8% | – |
| CritPt | 32.3% | 19.4% |
| GDPval (AA) | 54.4% | 43.4% |
| τ²-Bench Banking (AA) | 44.3% | – |
| ITBench SRE (AA) | 56.2% | – |
| Analyst Agent (AA) | 47.5% | – |
| BioMysteryBench | 71.1% | – |
| Code Migration | 52.9% | – |
| CUA-bench | 8.3% | – |
| Excel Modeling Benchmark | 72.3% | – |
| Finance Agent v2 | 53.8% | – |
| Harvey's Legal Agent Benchmark | 2.5% | – |
| Legal Research Bench | 48.1% | – |
| MedCode | 44.0% | – |
| MedScribe | 85.2% | – |
| MMMU-Pro (Vals) | 88.8% | – |
| MortgageTax | 67.3% | – |
| MysteryMechanism | 33.3% | – |
| ProgramBench | 1.5% | – |
| Public Benefits Bench | 66.5% | – |
| SAGE | 52.6% | – |
| SkillsBench | 54.1% | – |
| SREBench | 30.5% | – |
| Tax Agent Bench | 68.0% | – |
| Terminal-Bench 4.0 (Vals) | 27.8% | – |
| Terminal-Bench Science | 12.9% | – |
| Time Horizon Index: KSP | 23.8% | – |
| Vals Multimodal Index | 72.6% | – |
| Vibe Code Bench 1-100 | 20.0% | – |
| Vibe Code Bench v1.1 | 80.5% | – |
| Web Search Index | 43.6% | – |
| DrugDiscoveryBench | 56.5% | – |
| Chess Puzzles | 55.0% | – |
| EBR-bench | 44.8% | – |
| Mystery Game Puzzles | 58.0% | – |
| BALROG | 60.0% | – |
| PostTrainBench | 36.2% | – |
| DeepSWE | 72.7% | – |
| LMCA | 58.4% | – |
| DTBench | 95.5% | – |
| CursorBench | 41.7% | – |
| ALE-Bench | 2176.9 | – |
| GDP.pdf | 30.7% | – |
| FrontierSWE | 32.2% | – |
| BTF-3 | 13.7% | – |
| Terminal-Bench 4.0 (AA) | 39.9% | 12.6% |
| Terminal-Bench 2.1 (AA) | 88.0% | – |
| AutomationBench | 60.1% | 53.2% |
| GDP.pdf | 27.2% | 20.4% |
| MLCR | 26.1% | 16.1% |
| Harvey LAB | 87.2% | – |
| EnterpriseOps-Gym | 42.9% | – |
| AA-Omniscience: accuracy | 59.4% | 43.8% |
| AA-Omniscience: non-hallucination | 7.8% | 23.3% |
| AA-Briefcase | 1487 | 1299 |
Data as of 2026-09-23. Best configuration of each model; every score links to its source on the model pages.
GPT-5.6 Sol vs GPT-6 Luna: questions
- Is GPT-5.6 Sol better than GPT-6 Luna?
- GPT-5.6 Sol (max) leads on quality: 68.4 vs 61.2. The BenchLeader Index combines every independent quality benchmark; GPT-5.6 Sol (max) is ahead overall as of 2026-09-23, but check the category scores for your use.
- Which is cheaper, GPT-5.6 Sol or GPT-6 Luna?
- GPT-6 Luna is cheaper: $0.200 against $8.00 per million tokens, blended at three input tokens per output token.
- Which is faster, GPT-5.6 Sol or GPT-6 Luna?
- GPT-6 Luna streams faster: 157 against 85 output tokens per second.
- Which has the larger context window?
- Both accept 1.1M tokens of context.