Model comparison
Gemini 3 Pro vs GPT-5.6 Luna
Head-to-head evidence from 13 shared benchmark results across 5 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: Gemini 3 Pro #19 (Supported); GPT-5.6 Luna #22 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. Gemini 3 Pro and GPT-5.6 Luna share 13 comparable benchmark results. 2 of 8 categories are comparable. 14 results are unique to Gemini 3 Pro; 28 to GPT-5.6 Luna.
Updated July 23, 2026- Shared results
- 13
- Gemini 3 Pro only
- 14
- GPT-5.6 Luna only
- 28
- Comparable categories
- 2 / 8
Pick Gemini 3 Pro if you want the stronger benchmark profile. GPT-5.6 Luna only becomes the better choice if mathematics is the priority or you want the cheaper token bill.
Confidence note. This is a partial-evidence comparison with 13 shared benchmark results across 5 evidence categories; 2 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
Gemini 3 Pro has the cleaner BenchAlign overall profile here, landing at 67.73 versus 67.17. It is a real lead, but still close enough that category-level strengths matter more than the headline number.
Gemini 3 Pro's sharpest advantage is in multimodal & grounded, where it averages 81.1 against 78.4. The single biggest benchmark swing on the page is FrontierMath v2 (Tiers 1-3), 37.600% to 78.600%. GPT-5.6 Luna does hit back in mathematics, so the answer changes if that is the part of the workload you care about most.
Gemini 3 Pro is also the more expensive model on tokens at $2.00 input / $12.00 output per 1M tokens, versus $1.00 input / $6.00 output per 1M tokens for GPT-5.6 Luna. That is roughly 2.0x on output cost alone. GPT-5.6 Luna is the reasoning model in the pair, while Gemini 3 Pro is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. Gemini 3 Pro gives you the larger context window at 2M, compared with 1M for GPT-5.6 Luna.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | Gemini 3 Pro | Δ | GPT-5.6 Luna |
|---|---|---|---|
| Math | Gemini 3 Pro32.9 | Margin→ 40.7 | GPT-5.6 Luna73.6 |
| Multimodal | Gemini 3 Pro81.1 | Margin← 2.7 | GPT-5.6 Luna78.4 |
| Agentic | Gemini 3 ProNot measured | MarginNo overlap | GPT-5.6 Luna84.1 |
| Coding | Gemini 3 ProNot measured | MarginNo overlap | GPT-5.6 Luna62.7 |
| Reasoning | Gemini 3 Pro31.1 | MarginNo overlap | GPT-5.6 LunaNot measured |
| Knowledge | Gemini 3 ProNot measured | MarginNo overlap | GPT-5.6 Luna92.3 |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
FrontierMath v2 (Tiers 1-3)
MathA 37.600%B 78.600%Winner: GPT-5.6 LunaΔ 41FrontierMath v2 (Tiers 1-3): Gemini 3 Pro scored 37.600%; GPT-5.6 Luna scored 78.600%. GPT-5.6 Luna wins this benchmark. - Source ↗
FrontierMath v2 (Tier 4)
MathA 18.750%B 58.500%Winner: GPT-5.6 LunaΔ 39.8FrontierMath v2 (Tier 4): Gemini 3 Pro scored 18.750%; GPT-5.6 Luna scored 58.500%. GPT-5.6 Luna wins this benchmark. - Source ↗
MMMU-Pro
MultimodalA 81%B 78.4%Winner: Gemini 3 ProΔ 2.6MMMU-Pro: Gemini 3 Pro scored 81%; GPT-5.6 Luna scored 78.4%. Gemini 3 Pro wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | Gemini 3 Pro | GPT-5.6 Luna | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | Gemini 3 Pro$2 input / $12 output | GPT-5.6 Luna$1 input / $6 output | GPT-5.6 Luna has the lower combined listed price. |
| Generation speedtokens per second | Gemini 3 Pro109 tok/s | GPT-5.6 LunaNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | Gemini 3 Pro32.65 s | GPT-5.6 LunaNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | Gemini 3 Pro2M | GPT-5.6 Luna1M | Gemini 3 Pro lists the larger context window. |
Benchmark Deep Dive
Agentic18 benchmarks
| Benchmark | Gemini 3 Pro | GPT-5.6 Luna | Result |
|---|---|---|---|
| τ²-bench resultsSource | 87.1% | — | Not comparable |
| Gert LabsSource | 63.23% | — | Not comparable |
| JobBenchSource | 11.4% | — | Not comparable |
| Terminal-Bench 2.0Source | — | 84.7% | Not comparable |
| BrowseCompSource | — | 83.3% | Not comparable |
| OSWorld 2.0Source | — | 45.6% | Not comparable |
| CyberGymSource | — | 77.9% | Not comparable |
| ExploitGymSource | — | 12.4% | Not comparable |
| ToolathlonSource | — | 53.4% | Not comparable |
| AA Agentic IndexSource | — | 45.6% | Not comparable |
| GDPval-AASource | — | 54.2% | Not comparable |
| GDPval-AASource | — | 1584 | Not comparable |
| AA Harvey LABSource | — | 87.9% | Not comparable |
| AA ITBenchSource | — | 40.3% | Not comparable |
| AA Tau3 BankingSource | — | 27.2% | Not comparable |
| AA AutomationBenchSource | — | 42.2% | Not comparable |
| aaTerminalBench21Source | — | 80.9% | Not comparable |
| APEX-Agents-AASource | — | 35.8% | Not comparable |
Coding9 benchmarks
| Benchmark | Gemini 3 Pro | GPT-5.6 Luna | Result |
|---|---|---|---|
| Vibe Code BenchSource | 14.30% | — | Not comparable |
| AA-SciCodeSource | 56.1% | 52.5% | Gemini 3 Pro leads |
| AA LiveCodeBenchSource | 91.7% | — | Not comparable |
| SWE-bench ProSource | — | 62.7% | Not comparable |
| Terminal-Bench 2.0Source | — | 84.7% | Not comparable |
| deepSweSource | — | 67.2% | Not comparable |
| FrontierCode 1.1 ExtendedSource | — | 55.1% | Not comparable |
| cursorBench32Source | — | 61.1% | Not comparable |
| AA Coding IndexSource | — | 71.5% | Not comparable |
Reasoning4 benchmarks
Knowledge11 benchmarks
| Benchmark | Gemini 3 Pro | GPT-5.6 Luna | Result |
|---|---|---|---|
| Artificial Analysis Intelligence IndexSource | 39.5% | 51.2% | GPT-5.6 Luna leads |
| AA-GPQA DiamondSource | 90.8% | 91.1% | GPT-5.6 Luna leads |
| AA-HLESource | 37.2% | 37.2% | Tie |
| AA-Omniscience IndexSource | 15.8% | -11.2% | Gemini 3 Pro leads |
| AA-Omniscience AccuracySource | 55.9% | 41.5% | Gemini 3 Pro leads |
| AA-Omniscience Hallucination RateSource | 90.9% | 90.1% | GPT-5.6 Luna leads |
| AA MMLU-ProSource | 89.8% | — | Not comparable |
| GPQASource | — | 92.3% | Not comparable |
| GPQA-DSource | — | 92.3% | Not comparable |
| HealthBench ProfessionalSource | — | 55.7% | Not comparable |
| HealthBench HardSource | — | 32.0% | Not comparable |
MathGPT-5.6 Luna wins3 benchmarks
Multilingual1 benchmarks
| Benchmark | Gemini 3 Pro | GPT-5.6 Luna | Result |
|---|---|---|---|
| AA Global-MMLU-LiteSource | 92.2% | — | Not comparable |
MultimodalGemini 3 Pro wins8 benchmarks
| Benchmark | Gemini 3 Pro | GPT-5.6 Luna | Result |
|---|---|---|---|
| MMMU-ProSource | 81% | 78.4% | Gemini 3 Pro leads |
| MathVisionSource | 86.6% | — | Not comparable |
| VideoMMMUSource | 87.6% | — | Not comparable |
| ScreenSpot ProSource | 72.7% | — | Not comparable |
| CharXivSource | 81.4% | — | Not comparable |
| V*Source | 88.0% | — | Not comparable |
| AA-MMMU-ProSource | 80.2% | 78.6% | Gemini 3 Pro leads |
| MMMU-Pro w/ PythonSource | — | 79.5% | Not comparable |
Inst. Following1 benchmarks
| Benchmark | Gemini 3 Pro | GPT-5.6 Luna | Result |
|---|---|---|---|
| AA-IFBenchSource | 70.4% | — | Not comparable |
Frequently Asked Questions (3)
Which is better, Gemini 3 Pro or GPT-5.6 Luna?
Gemini 3 Pro is ahead on BenchLM's BenchAlign leaderboard, 67.73 to 67.17. The biggest single separator in this matchup is FrontierMath v2 (Tiers 1-3), where the scores are 37.600% and 78.600%.
Which is better for math, Gemini 3 Pro or GPT-5.6 Luna?
GPT-5.6 Luna has the edge for math in this comparison, averaging 73.6 versus 32.9. Inside this category, FrontierMath v2 (Tiers 1-3) is the benchmark that creates the most daylight between them.
Which is better for multimodal and grounded tasks, Gemini 3 Pro or GPT-5.6 Luna?
Gemini 3 Pro has the edge for multimodal and grounded tasks in this comparison, averaging 81.1 versus 78.4. Inside this category, MMMU-Pro is the benchmark that creates the most daylight between them.
Related Comparisons
Explore More
Choose a model with this week’s evidence
Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.
One email each week. Unsubscribe anytime.