Gemini 3 Pro vs GPT-5
Full profiles: Gemini 3 Pro (Google) and GPT-5 (OpenAI): every benchmark, price and rank.
Updated · Data from 17 sources
As of 16 Nov 2026, Gemini 3 Pro and GPT-5 are neck and neck on quality (60.0 vs 59.0 on the BenchLeader Index, inside the ±6.4 margin), and GPT-5 is 1.3× cheaper at $3.44 vs $4.50 per 1M tokens blended. Choose Gemini 3 Pro for reasoning (63 vs 48), multimodal (66 vs 57), and agents & tools (62 vs 58); choose GPT-5 for maths (65 vs 54) and instruction following (69 vs 61).
Key facts (as of 16 Nov 2026)
- Gemini 3 Pro scores 60.0 on the BenchLeader Index and GPT-5 scores 59.0 (BenchLeader, 16 Nov 2026).
- Gemini 3 Pro costs $4.50 per 1M tokens blended (list price, 16 Nov 2026).
- GPT-5 costs $3.44 per 1M tokens blended (list price, 16 Nov 2026).
- Of the 12 benchmarks both models have, Gemini 3 Pro leads 12 (BenchLeader, 16 Nov 2026).
- On Arena WebDev, Gemini 3 Pro is rated 1441 and GPT-5 1417 (LMArena, as published by 16 Nov 2026).
Which to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | Either | 58 vs 57 category score; Arena WebDev 1441 vs 1417 | Either |
| Agents & tool use | Gemini 3 Pro | 62 vs 58 category score; Terminal-Bench 69.4% vs 49.6% | Medium |
| Maths | GPT-5 | 65 vs 54 category score; OTIS Mock AIME 87.2% vs 91.4% | High |
| Reasoning | Gemini 3 Pro | 63 vs 48 category score; GPQA Diamond 92.6% vs 85.3% | High |
| Images and other media | Gemini 3 Pro | 66 vs 57 category score; GeoBench 84.0% vs 81.0% | High |
| Following instructions | GPT-5 | 69 vs 61 category score | High |
| Chat quality (human preference) | – | Only Gemini 3 Pro has an Arena rating (68 category score) | Not comparable |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Gemini 3 Pro and GPT-5 (medium) are neck and neck on quality (60.0 vs 59.0, within the ±6.4 margin).
- Gemini 3 Pro is stronger in agents & tools, coding, multimodal, reasoning.
- GPT-5 (medium) is stronger in instruction following, maths.
- GPT-5 (medium) is 1.3× cheaper ($3.44 vs $4.50 per 1M blended).
At a glance
| Spec | Gemini 3 Pro | GPT-5 |
|---|---|---|
| Developer | OpenAI | |
| Released | 18 Nov 2025 | 7 Aug 2025 |
| Context window (tokens) | – | 400,000 |
| Compared at | Gemini 3 Pro | GPT-5 (medium) |
| Open weights | No | No |
| Price in / out / cached input per 1M | – | $1.25 / $10.00 / $0.125 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Gemini 3 Pro | GPT-5 (medium) | Leader |
|---|---|---|---|
| Reasoning | 62.6 | 47.9 | Gemini 3 Pro |
| Coding | 57.8 | 57.4 | level |
| Agents & tools | 61.7 | 58.1 | Gemini 3 Pro |
| Maths | 54.0 | 65.4 | GPT-5 |
| Knowledge | 55.9 | – | not comparable |
| Instruction following | 61.1 | 69.4 | GPT-5 |
| Human preference | 67.7 | – | not comparable |
| Multimodal | 65.6 | 56.6 | Gemini 3 Pro |
| Long context | – | 70.4 | not comparable |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Not drawn, measured for fewer than two of these models: Knowledge, Long context, Human preference.
Where sources disagree
Across the 12 benchmarks both have, Gemini 3 Pro leads 12, GPT-5 leads 0. Every overall index that rates both puts Gemini 3 Pro ahead.
| Overall index | Gemini 3 Pro | GPT-5 (medium) | Leader |
|---|---|---|---|
| BenchLeader Index | 60.0 | 59.0 | neck and neck |
| Arena Text | 1486 | 1435† | Gemini 3 Pro |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | Gemini 3 Pro | GPT-5 (medium) |
|---|---|---|
| Blended price per 1M tokens | – | $3.44 |
| Input / output per 1M | – / – | $1.25 / $10.00 |
| Cached input per 1M | – | $0.125 |
| Price per task: chat reply | – | $0.0035 |
| Price per task: summarize a 30-page report | – | $0.021 |
| Price per task: code edit | – | $0.022 |
| Price per task: agentic coding session | – | $0.115 |
| Price per task: structured extraction | – | $0.0045 |
| Agentic coding session, input mostly cached | – | $0.064 |
| Output speed | – | 82 tok/s |
| Time to first answer | – | 22.9 s |
| Time to a complete answer, last measured 13 Nov 2026 | – | 66.2 s |
Each model is compared at its best-ranked reasoning-effort setting. Gemini 3 Pro scores 60.0; Gemini 3 Pro (high) scores 60.0. GPT-5 (medium) scores 59.0 and takes 23 s to its first answer token; GPT-5 (high) scores 58.2 in 49 s, GPT-5 scores 56.6 in 5 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
128 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Gemini 3 Pro | GPT-5 (medium) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 60.0 | 59.0 |
| Agents & tools score | BenchLeader | 61.7 | 58.1 |
| Coding score | BenchLeader | 57.8 | 57.4 |
| Human preference score | BenchLeader | 67.7 | – |
| Instruction following score | BenchLeader | 61.1 | 69.4 |
| Knowledge score | BenchLeader | 55.9 | – |
| Maths score | BenchLeader | 54.0 | 65.4 |
| Multimodal score | BenchLeader | 65.6 | 56.6 |
| Reasoning score | BenchLeader | 62.6 | 47.9 |
| Long context score | BenchLeader | – | 70.4 |
| Blended price $/M | BenchLeader | – | $3.44 |
| Output speed | BenchLeader | – | 82 tok/s |
| Time to first answer | BenchLeader | – | 22.9 s |
| Context window | BenchLeader | – | 400,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 92.6% | 85.3% |
| Humanity's Last Exam | Scale AI / CAIS | 37.5% | 25.3%† |
| SimpleBench | SimpleBench | 76.4% | 56.7%† |
| Arena Hard Prompts | Arena (LMArena) | 1502 | 1449† |
| GPQA Diamond (AA) | Artificial Analysis | 90.8%† | 84.2% |
| Humanity's Last Exam (AA) | Artificial Analysis | 39.7%† | 25.4% |
| GPQA Diamond (Vals) | Vals AI | 91.7%† | 85.6%† |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 80.1% | 72.7%† |
| GPQA Diamond (HELM) | HELM Capabilities | 80.3% | 79.1%† |
| ARC-AGI-1 | ARC Prize | 75.0% | 56.2% |
| ARC-AGI-2 | ARC Prize | 31.1% | 7.5% |
| CritPt | Artificial Analysis | 9.1%† | 0.0% |
| CritPt v2 hard subset (AA) | Artificial Analysis | – | 11.5%† |
| SciPredict | Scale AI SEAL | 25.3% | – |
| Arena Expert | Arena (LMArena) | 1499 | 1445† |
| Chess Puzzles | Epoch AI Benchmarking Hub | – | 32.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | – | 15.0% |
| CL-bench | Epoch AI Benchmarking Hub | 15.8% | – |
| LMCA | Epoch AI Benchmarking Hub | – | 40.0%† |
| DTBench | Epoch AI Benchmarking Hub | – | 90.7%† |
| ForecastBench | ForecastBench | 65.0% | 65.0%† |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 72.9% | 71.5% |
| SciCode | SciCode | – | 42.9%† |
| WeirdML | WeirdML | 69.9% | 39.8%† |
| GSO-Bench | GSO-Bench | 18.6% | 6.9%† |
Show all 128 benchmarks
| Metric | Publisher | Gemini 3 Pro | GPT-5 (medium) |
|---|---|---|---|
| Coding | |||
| Arena Coding | Arena (LMArena) | 1517 | 1462† |
| Arena WebDev | Arena (LMArena) | 1441 | 1417 |
| LiveCodeBench | Vals AI | 86.4% | 85.9%† |
| SWE-bench (Vals) | Vals AI | 76.4%† | 69.0%† |
| SWE-Bench Pro | Scale AI SEAL | 43.3% | 41.8%† |
| Aider Polyglot | Aider polyglot leaderboard | – | 86.7% |
| SWE-bench Verified (bash only) | SWE-bench | 74.2% | 65.0% |
| SWE-bench Multilingual (bash only) | SWE-bench | 68.7% | – |
| SWE-bench Verified (any scaffold) | SWE-bench | 77.4% | 74.4%† |
| Vibe Code Bench v1.1 | Vals AI | 14.3%† | 20.1%† |
| SWE-Bench Pro (private) | Scale AI SEAL | 17.9% | 14.9%† |
| MirrorCode | Epoch AI Benchmarking Hub | – | 4.4%† |
| ALE-Bench | Epoch AI Benchmarking Hub | 1176.8 | 1162.5† |
| AlgoTune | AlgoTune | 1.8 | 1.7† |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | – | 35.2%† |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | – | 38.0 |
| Design Arena Website | Design Arena | 1244 | 1174† |
| LiveCodeBench (AA) | Artificial Analysis | 91.8%† | 70.3% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 69.4% | 49.6% |
| GDPval | OpenAI | 40.3% | 34.8% |
| Remote Labor Index | Scale AI / CAIS | 1.3% | 1.7%† |
| Terminal-Bench Hard | Artificial Analysis | 41.7%† | 37.9% |
| τ²-Bench Telecom (AA) | Artificial Analysis | 87.1%† | 86.5% |
| MCP Atlas | Scale AI SEAL | 70.3% | – |
| BFCL Overall | Berkeley Function Calling Leaderboard | 72.5% | – |
| τ²-bench | τ²-bench | 82.5%† | – |
| τ³-bench Banking | τ²-bench | 18.0%† | – |
| GDPval-AA v2.1 | Artificial Analysis | – | 925† |
| τ³-Banking (AA) | Artificial Analysis | – | 22.1%† |
| Poker Agent | Vals AI | 1078.9% | 1103.2%† |
| Terminal-Bench 2.0 (Vals) | Vals AI | 55.1%† | 37.1%† |
| Arena Search | Arena (LMArena) | 1201 | 1189† |
| BALROG | BALROG | 58.1% | 32.8%† |
| DeepResearch Bench | DeepResearch Bench | 46.3%† | 48.6% |
| METR Time Horizons | METR | 71.0% | 69.6% |
| Vending-Bench 2 | Andon Labs | 5478.2 | – |
| Design Arena Agentic | Design Arena | 1080 | – |
| BrowseComp-Plus | BrowseComp-Plus | – | 57.6%† |
| BrowseComp | OpenAI | – | 20.1%† |
| DeepSearchQA | – | 59.4%† | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | – | 55.4%† |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | – | 21.9%† |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 91.4% | 87.2% |
| MATH Level 5 | Epoch AI Benchmarking Hub | – | 97.9% |
| ProofBench | Vals AI | 20.0% | 18.0%† |
| AIME (Vals) | Vals AI | 96.7%† | 93.4%† |
| MGSM | Vals AI | 93.9%† | 92.8%† |
| AIME 2026 | MathArena | 91.7% | – |
| HMMT February 2026 | MathArena | 86.4% | – |
| IMO 2025 | MathArena | – | 38.1%† |
| MathArena Apex | MathArena | 23.4% | 1.0%† |
| Omni-MATH (HELM) | HELM Capabilities | 55.6% | 64.7%† |
| MATH 500 | Vals AI | 96.4%† | 96.0%† |
| Arena Maths | Arena (LMArena) | 1475 | 1411† |
| AIME 2025 (AA) | Artificial Analysis | 95.7%† | 91.7% |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | – | 50.1%† |
| MMLU-Pro | Vals AI | 90.1% | 86.5%† |
| LegalBench | Vals AI | 87.0%† | 86.0%† |
| CorpFin | Vals AI | 63.7%† | 61.1%† |
| TaxEval | Vals AI | 72.6%† | 73.4%† |
| MedQA | Vals AI | 96.0%† | 96.3%† |
| PRBench Finance | Scale AI SEAL | 39.2% | 51.3%† |
| PRBench Legal | Scale AI SEAL | 40.6% | 49.0%† |
| MultiNRC | Scale AI SEAL | 59.0% | 52.1%† |
| MMLU-Pro (HELM) | HELM Capabilities | 90.3% | 86.3%† |
| CaseLaw v2 | Vals AI | 53.1%† | 66.5%† |
| MedCode | Vals AI | 52.2%† | 49.6%† |
| MedScribe | Vals AI | 72.0%† | 83.7%† |
| EBR-bench | Epoch AI Benchmarking Hub | – | 12.7%† |
| Global-MMLU-Lite (AA) | Artificial Analysis | 92.2%† | 90.4% |
| Instruction following | |||
| IFBench | Artificial Analysis | 70.4%† | 70.6% |
| MultiChallenge | Scale AI SEAL | 65.7% | 63.2%† |
| TutorBench | Scale AI SEAL | 53.7% | 55.3%† |
| IFEval (HELM) | HELM Capabilities | 87.6% | 87.5%† |
| WildBench (HELM) | HELM Capabilities | 85.9% | 85.7%† |
| Arena Instruction Following | Arena (LMArena) | 1473 | 1415† |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1486 | 1427† |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1524 | 1626† |
| Arena Creative Writing | Arena (LMArena) | 1484 | 1389† |
| Arena Multi-turn | Arena (LMArena) | 1495 | 1449† |
| Arena Longer Queries | Arena (LMArena) | 1491 | 1435† |
| Lech Mazur Creative Writing | Lech Mazur | – | 86.0% |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1305 | 1232† |
| MMMU-Pro | Artificial Analysis | 80.2%† | 74.3% |
| VISTA | Scale AI SEAL | 51.5% | 49.7%† |
| MMMU-Pro (Vals) | Vals AI | 87.5%† | 81.5%† |
| MortgageTax | Vals AI | 69.1%† | 65.5%† |
| AudioMC | Scale AI SEAL | 54.6%† | – |
| AudioMC (text output) | Scale AI SEAL | 54.6%† | – |
| VTB | Scale AI SEAL | 26.9% | 17.0%† |
| GeoBench | GeoBench | 84.0% | 81.0% |
| VPCT | Epoch AI Benchmarking Hub | 91.0% | 63.2% |
| Long context | |||
| Fiction.LiveBench 120k | Fiction.live | – | 96.9% |
| AA-LCR v1.1 | Artificial Analysis | 76.0%† | 76.0% |
| Arena Document | Arena (LMArena) | 1451 | – |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 153.0 | 150† |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 28.0† | 22.9 |
| AA Progress Index v1 | Artificial Analysis | 290.7† | 264.4 |
| HELM Capabilities mean | HELM Capabilities | 79.9% | 80.7%† |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 41.7% | 17.0%† |
| MASK | Scale AI SEAL | 42.6% | 79.3%† |
| PropensityBench | Scale AI SEAL | 52.9% | – |
| Metadata | |||
| AA Openness Index | Artificial Analysis | 5.6† | 5.6 |
Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- GPT-5 (medium): BenchLeader Index 58.2 → 59.0.
- Gemini 3 Pro: 1 score revision.
- GPT-5 (medium): 1 new result (Aider Polyglot).
- GPT-5 (medium): 2 new results (Global-MMLU-Lite (AA), AIME 2025 (AA)); 1 score revision.
- GPT-5 (medium): 1 score revision.
- Gemini 3 Pro: 1 new result (Arena Search).
- GPT-5 (medium): 1 score revision.
- GPT-5 (medium): 1 score revision.
- GPT-5 (medium): 1 new result (AA Progress Index v1).
- Gemini 3 Pro: 1 score revision.
Related comparisons
- Gemini 3 Pro vs Gemini 4 Carbon
- Gemini 3 Pro vs Claude Opus 5.5
- Gemini 3 Pro vs GPT-6 Astra
- GPT-5 vs Claude Opus 5.5
- Gemini 3 Pro vs GPT-6.1 Sol
- Gemini 3 Pro vs Gemini 4 Argon
- Gemini 3 Pro vs Claude Sonnet 5.5
- Gemini 3 Pro vs GPT-5.5
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Gemini 3 Pro vs GPT-5: questions
- Is Gemini 3 Pro better than GPT-5?
- Gemini 3 Pro and GPT-5 (medium) are neck and neck on quality (60.0 vs 59.0, within the ±6.4 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
- Gemini 3 Pro vs GPT-5: which is cheaper, and will that change?
- GPT-5 is cheaper today: $3.44 against $4.50 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Gemini 3 Pro or GPT-5 better for coding?
- They are level in coding (58 vs 57 on the category index). On Arena WebDev, the top shared coding benchmark, it is 1441 for Gemini 3 Pro against 1417 for GPT-5.
- Is Gemini 3 Pro or GPT-5 better for agents and tool use?
- Gemini 3 Pro scores higher in agentic tasks (62 vs 58 on the category index, where 50 is average). On Terminal-Bench, the top shared agents & tools benchmark, it is 69.4% for Gemini 3 Pro against 49.6% for GPT-5.
- Do independent benchmarks agree on Gemini 3 Pro vs GPT-5?
- Across the 12 benchmarks both have, Gemini 3 Pro leads 12, GPT-5 leads 0. Every overall index that rates both puts Gemini 3 Pro ahead.
More questions (1)
- When was Gemini 3 Pro released?
- Gemini 3 Pro was released on 18 Nov 2025.