GPT-5.6 Luna
OpenAI·Released Jul 9, 2026
Updated Oct 8, 11:55 PM ET
Overall score
50.4
Rank #36 overall
Evidence
34 results
97% of scored kinds covered
Context window
1.1M tokens
From OpenRouter
API price, USD per 1M tokens
$0.20 / $1.20
Input / output · cached input $0.02
The score is on the board's scale (reference average 50, about 15 points per standard deviation), not a percentage.
By domain
Each domain is scored on the same scale as the overall score. Coverage is the share of the domain's kinds of evaluation with a result; it says how much evidence there is, not how good the model is.
Every result behind the score
The result as the evaluator publishes it, and the reading it gives on the board's scale. A hard evaluation can give a high reading for a modest result; results near 95% or 5% only say "at least" or "at most".
Coding
10 results · 55.4- Vibe Code Bench 1–100Vals AI · max effort22.6%Reading 70.3
- ProgramBenchVals AI · max effortNear the floor0.0%Reading at most 65.7
- Code MigrationVals AI · max effort44.5%Reading 63.9
- CyberBench PatchVals AI · max effort83.9%Reading 62.4
- Vibe Code BenchVals AI · max effort77.1%Reading 61.8
- LiveBench CodingLiveBench · max effort65.7%Reading 55.7
- SciCodeArtificial Analysis · high effort51.6%Reading 54.0
- Terminal-Bench 4.0Vals AI · max effort11.6%Reading 47.0
- APEX-SWEMercor · max effort32.1%Reading 40.2
- Terminal-Bench 4.0 (AA run)Artificial Analysis · high effortNear the floor2.5%Reading at most 34.5
Research and reasoning
9 results · 52.2- FrontierMath Tiers 1–3Epoch AI · max effort82.1%Reading 69.3
- FrontierMath Tier 4Epoch AI · max effort61.0%Reading 67.5
- ProofBenchVals AI · max effort60.0%Reading 60.6
- Chess PuzzlesEpoch AI · max effort40.0%Reading 60.0
- LiveBench ReasoningLiveBench · max effort86.4%Reading 52.1
- Terminal-Bench ScienceVals AI · max effortNear the floor0.0%Reading at most 49.0
- Humanity's Last Exam (AA run)Artificial Analysis · high effort33.4%Reading 47.4
- MysteryMechanismVals AI · max effort14.4%Reading 42.2
- Mystery Game PuzzlesEpoch AI · max effort21.0%Reading 39.3
Professional work
9 results · 51.8- LiveBench Data AnalysisLiveBench · max effort78.0%Reading 67.5
- EMBVals AI · max effort67.1%Reading 65.3
- Finance AgentVals AI · max effort55.0%Reading 63.5
- Legal Research BenchVals AI · max effort36.5%Reading 58.0
- Tax Agent BenchVals AI · max effort24.8%Reading 55.2
- APEX-AgentsMercor · max effort43.0%Reading 48.6
- MedCodeVals AI · max effort42.4%Reading 42.2
- τ-Bench Banking (AA run)Artificial Analysis · high effort25.2%Reading 41.7
- Harvey Legal Agent BenchmarkVals AI · max effortNear the floor1.3%Reading at most 24.8
Knowledge and accuracy
5 results · 34.8- AA-LCRArtificial Analysis · high effort80.3%Reading 49.7
- SimpleQA VerifiedEpoch AI · max effort41.0%Reading 41.8
- LiveBench LanguageLiveBench · max effort72.6%Reading 39.1
- LiveBench Instruction FollowingLiveBench · max effort60.1%Reading 24.5
- BullshitBenchBullshitBench · max effortCapped18.2%Reading -14.4
Human preference
1 result · 44.2- Arena TextLMArena · xhigh effort1,453.1Reading 41.4
Shown, not scored
Composite indices, saturated or older evaluations, evaluations still being checked, and vision, writing and multilingual results. They do not move the score.
- IOIVals AI · CodingReference only61.8%
- SkillsBenchVals AI · CodingWatching60.4%
- SRE BenchVals AI · CodingWatching1.5%
- SWE-bench VerifiedVals AI · CodingReference only93.0%
- BioMysteryBenchVals AI · Research and reasoningWatching61.5%
- OTIS Mock AIMEEpoch AI · Research and reasoningReference only98.3%
- APEX-AccountingMercor · Professional workWatching4.4%
- CorpFinVals AI · Professional workReference only64.2%
- LegalBenchVals AI · Professional workReference only84.0%
- MedScribeVals AI · Professional workWatching84.4%
- TaxEvalVals AI · Professional workReference only76.2%
- GPQA DiamondEpoch AI · Knowledge and accuracyReference only91.6%
- GPQA DiamondVals AI · Knowledge and accuracyReference only91.7%
- MMLU-ProVals AI · Knowledge and accuracyReference only86.0%
- AA Intelligence IndexArtificial Analysis · Composite indicesReference only32.1
- LiveBench averageLiveBench · Composite indicesReference only73.6%
- Vals IndexVals AI · Composite indicesReference only51.7%
- Arena VisionLMArena · VisionReference only1,259.8
- MMMU ProVals AI · VisionReference only85.0%
No result yet
Scored evaluations this model has not taken. A missing result neither adds nor subtracts.
- τ²-Bench Telecom (AA run) · Artificial Analysis
- IFBench (AA run) · Artificial Analysis