DeepSeek V4 Pro vs Gemini 3 Pro
Full profiles: DeepSeek V4 Pro (DeepSeek) and Gemini 3 Pro (Google): every benchmark, price and rank.
Updated · Data from 16 sources
As of 16 Nov 2026, DeepSeek V4 Pro and Gemini 3 Pro are neck and neck on quality (56.0 vs 60.0 on the BenchLeader Index, inside the ±5.6 margin), and DeepSeek V4 Pro is 4.5× cheaper at $0.990 vs $4.50 per 1M tokens blended. Choose DeepSeek V4 Pro for maths (65 vs 54); choose Gemini 3 Pro for knowledge (56 vs 41), reasoning (63 vs 56), and coding (58 vs 52).
Key facts (as of 16 Nov 2026)
- DeepSeek V4 Pro scores 56.0 on the BenchLeader Index and Gemini 3 Pro scores 60.0 (BenchLeader, 16 Nov 2026).
- DeepSeek V4 Pro costs $0.990 per 1M tokens blended (list price, 16 Nov 2026).
- Gemini 3 Pro costs $4.50 per 1M tokens blended (list price, 16 Nov 2026).
- Of the 15 benchmarks both models have, Gemini 3 Pro leads 11 (BenchLeader, 16 Nov 2026).
- On Arena WebDev, DeepSeek V4 Pro is rated 1581 and Gemini 3 Pro 1441 (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 | Gemini 3 Pro | 58 vs 52 category score; WeirdML 69.9% vs 46.5% | Medium |
| Maths | DeepSeek V4 Pro | 65 vs 54 category score; OTIS Mock AIME 95.6% vs 91.4% | High |
| Reasoning | Gemini 3 Pro | 63 vs 56 category score; GPQA Diamond 92.6% vs 90.9% | Medium |
| Knowledge | Gemini 3 Pro | 56 vs 41 category score | High |
| Chat quality (human preference) | Gemini 3 Pro | 68 vs 65 on Arena votes | Medium |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- DeepSeek V4 Pro (high) and Gemini 3 Pro are neck and neck on quality (56.0 vs 60.0, within the ±5.6 margin).
- DeepSeek V4 Pro (high) is stronger in maths.
- Gemini 3 Pro is stronger in coding, human preference, knowledge, reasoning.
- DeepSeek V4 Pro (high) is 4.5× cheaper ($0.990 vs $4.50 per 1M blended).
At a glance
| Spec | DeepSeek V4 Pro | Gemini 3 Pro |
|---|---|---|
| Developer | DeepSeek | |
| Released | 13 Aug 2026 | 18 Nov 2025 |
| Context window (tokens) | 1,048,576 | – |
| Compared at | DeepSeek V4 Pro (high) | Gemini 3 Pro |
| Open weights | Yes | No |
| Price in / out / cached input per 1M | $1.32 / $3.96 / $0.022 | – |
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 | DeepSeek V4 Pro (high) | Gemini 3 Pro | Leader |
|---|---|---|---|
| Reasoning | 55.6 | 62.6 | Gemini 3 Pro |
| Coding | 51.5 | 57.8 | Gemini 3 Pro |
| Agents & tools | – | 61.7 | not comparable |
| Maths | 64.5 | 54.0 | DeepSeek V4 Pro |
| Knowledge | 41.2 | 55.9 | Gemini 3 Pro |
| Instruction following | – | 61.1 | not comparable |
| Human preference | 65.0 | 67.7 | Gemini 3 Pro |
| Multimodal | – | 65.6 | not comparable |
| Long context | 64.3 | – | 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: Agents & tools, Instruction following, Long context, Multimodal.
Where sources disagree
Across the 15 benchmarks both have, DeepSeek V4 Pro leads 4, Gemini 3 Pro leads 11. Every overall index that rates both puts Gemini 3 Pro ahead.
| Overall index | DeepSeek V4 Pro (high) | Gemini 3 Pro | Leader |
|---|---|---|---|
| BenchLeader Index | 56.0 | 60.0 | neck and neck |
| Arena Text | 1464 | 1486 | 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 | DeepSeek V4 Pro (high) | Gemini 3 Pro |
|---|---|---|
| Blended price per 1M tokens | $1.98 | – |
| Input / output per 1M | $1.32 / $3.96 | – / – |
| Cached input per 1M | $0.022 | – |
| Price per task: chat reply | $0.0017 | – |
| Price per task: summarize a 30-page report | $0.018 | – |
| Price per task: code edit | $0.014 | – |
| Price per task: agentic coding session | $0.095 | – |
| Price per task: structured extraction | $0.0034 | – |
| Agentic coding session, input mostly cached | $0.037 | – |
| Cost per benchmark task (AA Index) | $0.849 | – |
| Output tokens per task (AA Index) | 70k | – |
| Of which reasoning | 45k | – |
| Time per task (AA Index) | 15 min | – |
| Output speed | 80 tok/s | – |
| Time to first answer | 42.8 s | – |
Each model is compared at its best-ranked reasoning-effort setting. DeepSeek V4 Pro (high) scores 56.0 and takes 43 s to its first answer token; DeepSeek V4 Pro (max) scores 54.7 in 47 s, DeepSeek V4 Pro scores 53.0 in 1 s. Gemini 3 Pro scores 60.0; Gemini 3 Pro (high) scores 60.0. 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
180 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 | DeepSeek V4 Pro (high) | Gemini 3 Pro |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 56.0 | 60.0 |
| Coding score | BenchLeader | 51.5 | 57.8 |
| Human preference score | BenchLeader | 65.0 | 67.7 |
| Knowledge score | BenchLeader | 41.2 | 55.9 |
| Long context score | BenchLeader | 64.3 | – |
| Maths score | BenchLeader | 64.5 | 54.0 |
| Reasoning score | BenchLeader | 55.6 | 62.6 |
| Agents & tools score | BenchLeader | – | 61.7 |
| Instruction following score | BenchLeader | – | 61.1 |
| Multimodal score | BenchLeader | – | 65.6 |
| Blended price $/M | BenchLeader | $1.98 | – |
| Output speed | BenchLeader | 80 tok/s | – |
| Time to first answer | BenchLeader | 42.8 s | – |
| Context window | BenchLeader | 1,048,576 tokens | – |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 90.9% | 92.6% |
| Humanity's Last Exam | Scale AI / CAIS | – | 37.5% |
| SimpleBench | SimpleBench | – | 76.4% |
| Arena Hard Prompts | Arena (LMArena) | 1482 | 1502 |
| LiveBench 1.0 Reasoning | LiveBench | 85.8%† | – |
| GPQA Diamond (AA) | Artificial Analysis | 92.8%† | 90.8%† |
| Humanity's Last Exam (AA) | Artificial Analysis | 41.5% | 39.7%† |
| HLE-Diamond (AA) | Artificial Analysis | 20.5% | – |
| GPQA Diamond (Vals) | Vals AI | 92.4%† | 91.7%† |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 53.5%† | 80.1% |
| GPQA Diamond (HELM) | HELM Capabilities | – | 80.3% |
| ARC-AGI-1 | ARC Prize | 87.2% | 75.0% |
| ARC-AGI-2 | ARC Prize | 59.7% | 31.1% |
| CritPt | Artificial Analysis | 18.6% | 9.1%† |
| CritPt v2 hard subset (AA) | Artificial Analysis | 28.5% | – |
| SciPredict | Scale AI SEAL | – | 25.3% |
| HLE-Diamond (Scale) | Scale AI SEAL | 13.4%† | – |
| Arena Expert | Arena (LMArena) | 1481 | 1499 |
| Chess Puzzles | Epoch AI Benchmarking Hub | 16.0% | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 43.0%† | – |
| CL-bench | Epoch AI Benchmarking Hub | – | 15.8% |
| CL-bench Life | Epoch AI Benchmarking Hub | 13.5% | – |
| LMCA | Epoch AI Benchmarking Hub | 45.5%† | – |
| DTBench | Epoch AI Benchmarking Hub | 93.9%† | – |
| ForecastBench | ForecastBench | 60.7%† | 65.0% |
Show all 180 benchmarks
| Metric | Publisher | DeepSeek V4 Pro (high) | Gemini 3 Pro |
|---|---|---|---|
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 77.6%† | 72.9% |
| SciCode | SciCode | 46.4% | – |
| WeirdML | WeirdML | 46.5% | 69.9% |
| FrontierCode | Cognition | 28.6% | – |
| GSO-Bench | GSO-Bench | – | 18.6% |
| Arena Coding | Arena (LMArena) | 1503 | 1517 |
| Arena WebDev | Arena (LMArena) | 1581 | 1441 |
| LiveBench 1.0 Coding | LiveBench | 77.2%† | – |
| SciCode (AA) | Artificial Analysis | 51.7% | – |
| LiveCodeBench | Vals AI | 87.5%† | 86.4% |
| IOI | Vals AI | 51.6%† | – |
| SWE-bench (Vals) | Vals AI | 96.4%† | 76.4%† |
| SWE-Bench Pro | Scale AI SEAL | – | 43.3% |
| SWE-bench Verified (bash only) | SWE-bench | – | 74.2% |
| SWE-bench Multilingual (bash only) | SWE-bench | – | 68.7% |
| SWE-bench Verified (any scaffold) | SWE-bench | – | 77.4% |
| Code Migration | Vals AI | 41.5%† | – |
| Code Migration v1.1 | Vals AI | 52.0%† | – |
| ProgramBench | Vals AI | 0.0%† | – |
| Vibe Code Bench 1-100 | Vals AI | 17.5%† | – |
| Vibe Code Bench v1.1 | Vals AI | 82.3%† | 14.3%† |
| SWE-Bench Pro (private) | Scale AI SEAL | – | 17.9% |
| SWE Atlas v1.0: Codebase QnA | Scale AI SEAL | 27.1%† | – |
| SWE Atlas v1.0: Test Writing | Scale AI SEAL | 27.1%† | – |
| MirrorCode | Epoch AI Benchmarking Hub | 33.3%† | – |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 40.0% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 1006.1 | 1176.8 |
| AlgoTune | AlgoTune | – | 1.8 |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 6.1% | – |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 78.7%† | – |
| Startrise Frontend | Startrise | 62.7%† | – |
| GameBench 2 Overall | DuelLab GameBench 2 | 41.9† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 45.1 | – |
| Design Arena Website | Design Arena | 1227† | 1244 |
| FrontierCode 1.1 (AA) | Artificial Analysis | 38.8%† | – |
| FrontierCode 2.0 (AA) | Artificial Analysis | 16.7% | – |
| LiveCodeBench (AA) | Artificial Analysis | – | 91.8%† |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 6.7%† | 69.4% |
| GDPval | OpenAI | – | 40.3% |
| Remote Labor Index | Scale AI / CAIS | – | 1.3% |
| APEX-Agents | Mercor | 47.3%† | – |
| LiveBench Agentic Coding | LiveBench | 55.0%† | – |
| Terminal-Bench Hard | Artificial Analysis | – | 41.7%† |
| τ²-Bench Telecom (AA) | Artificial Analysis | – | 87.1%† |
| Terminal-Bench 2.1 (Vals) | Vals AI | 50.2%† | – |
| MCP Atlas | Scale AI SEAL | – | 70.3% |
| HiL-Bench | Scale AI SEAL | 38.5%† | – |
| 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, normalised score | Artificial Analysis | 47.5%† | – |
| GDPval-AA v2.1 | Artificial Analysis | 1336 | – |
| τ³-Banking (AA) | Artificial Analysis | 39.6%† | – |
| CyberBench | Vals AI | 72.9%† | – |
| Finance Agent v2 | Vals AI | 50.4%† | – |
| Finance Agent v2.1 | Vals AI | 57.1%† | – |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 10.8%† | – |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 7.5%† | – |
| Legal Research Bench | Vals AI | 40.9%† | – |
| Poker Agent | Vals AI | – | 1078.9% |
| SkillsBench | Vals AI | 53.8%† | – |
| SREBench | Vals AI | 1.9%† | – |
| Tax Agent Bench | Vals AI | 26.8%† | – |
| Terminal-Bench 2.0 (Vals) | Vals AI | 56.2%† | 55.1%† |
| Terminal-Bench 4.0 (Vals) | Vals AI | 14.1%† | – |
| Terminal-Bench Science | Vals AI | 4.3%† | – |
| Arena Search | Arena (LMArena) | – | 1201 |
| BALROG | BALROG | – | 58.1% |
| DeepResearch Bench | DeepResearch Bench | – | 46.3%† |
| METR Time Horizons | METR | – | 71.0% |
| Vending-Bench 2 | Andon Labs | 3284.5† | 5478.2 |
| AutomationBench | Artificial Analysis | 46.2% | – |
| GDP.pdf (AA) | Artificial Analysis | 10.4% | – |
| EnterpriseOps-Gym | Artificial Analysis | 49.6%† | – |
| AA-Briefcase v1.1 | Artificial Analysis | 1148 | – |
| Design Arena Agentic | Design Arena | 927† | 1080 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 1.9% | – |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 64.6%† | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 26.8%† | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 95.6% | 91.4% |
| ProofBench | Vals AI | 50.0%† | 20.0% |
| LiveBench Mathematics | LiveBench | 95.1%† | – |
| AIME (Vals) | Vals AI | – | 96.7%† |
| MGSM | Vals AI | – | 93.9%† |
| AIME 2026 | MathArena | 96.7%† | 91.7% |
| HMMT February 2026 | MathArena | 93.9%† | 86.4% |
| USAMO 2026 | MathArena | 60.7%† | – |
| MathArena Apex | MathArena | 28.1%† | 23.4% |
| Omni-MATH (HELM) | HELM Capabilities | – | 55.6% |
| MATH 500 | Vals AI | – | 96.4%† |
| Arena Maths | Arena (LMArena) | 1465 | 1475 |
| AIME 2025 (AA) | Artificial Analysis | – | 95.7%† |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 52.9%† | – |
| LiveBench 1.0 Data Analysis | LiveBench | 79.2%† | – |
| AA-Omniscience v1.1 | Artificial Analysis | -6.5 | – |
| MMLU-Pro | Vals AI | 87.3%† | 90.1% |
| LegalBench | Vals AI | 82.4%† | 87.0%† |
| CorpFin | Vals AI | 65.4%† | 63.7%† |
| TaxEval | Vals AI | 73.1%† | 72.6%† |
| MedQA | Vals AI | – | 96.0%† |
| PRBench Finance | Scale AI SEAL | – | 39.2% |
| PRBench Legal | Scale AI SEAL | – | 40.6% |
| MultiNRC | Scale AI SEAL | – | 59.0% |
| MMLU-Pro (HELM) | HELM Capabilities | – | 90.3% |
| CaseLaw v2 | Vals AI | 59.4%† | 53.1%† |
| Excel Modeling Benchmark | Vals AI | 52.8%† | – |
| Excel Modeling Benchmark v1.1 | Vals AI | 65.1%† | – |
| MedCode | Vals AI | 42.5%† | 52.2%† |
| MedScribe | Vals AI | 80.2%† | 72.0%† |
| Public Benefits Bench v1.1 | Vals AI | 62.9%† | – |
| Immigration Bench | Vals AI | 61.7%† | – |
| DrugDiscoveryBench | Scale AI SEAL | 31.7%† | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 45.1% | – |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 5.9% | – |
| Global-MMLU-Lite (AA) | Artificial Analysis | – | 92.2%† |
| Instruction following | |||
| LiveBench Language | LiveBench | 82.1%† | – |
| LiveBench Instruction Following | LiveBench | 67.7%† | – |
| IFBench | Artificial Analysis | – | 70.4%† |
| MultiChallenge | Scale AI SEAL | – | 65.7% |
| TutorBench | Scale AI SEAL | – | 53.7% |
| IFEval (HELM) | HELM Capabilities | – | 87.6% |
| WildBench (HELM) | HELM Capabilities | – | 85.9% |
| Arena Instruction Following | Arena (LMArena) | 1461 | 1473 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1464 | 1486 |
| EQ-Bench 4 | EQ-Bench | 1156† | – |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1553† | 1524 |
| Arena Creative Writing | Arena (LMArena) | 1444 | 1484 |
| Arena Multi-turn | Arena (LMArena) | 1469 | 1495 |
| Arena Longer Queries | Arena (LMArena) | 1474 | 1491 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | – | 1305 |
| MMMU-Pro | Artificial Analysis | – | 80.2%† |
| VISTA | Scale AI SEAL | – | 51.5% |
| MMMU-Pro (Vals) | Vals AI | – | 87.5%† |
| MortgageTax | Vals AI | – | 69.1%† |
| AudioMC | Scale AI SEAL | – | 54.6%† |
| AudioMC (text output) | Scale AI SEAL | – | 54.6%† |
| VTB | Scale AI SEAL | – | 26.9% |
| GeoBench | GeoBench | – | 84.0% |
| VPCT | Epoch AI Benchmarking Hub | – | 91.0% |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 79.3% | 76.0%† |
| Arena Document | Arena (LMArena) | – | 1451 |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 155.9† | 153.0 |
| LiveBench 1.0 | LiveBench | 77.4%† | – |
| AA Intelligence Index v5.0.1 | Artificial Analysis | 20.8 | – |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 36† | 28.0† |
| AA Progress Index v1 | Artificial Analysis | 311.5† | 290.7† |
| Vals Index | Vals AI | 53.0† | – |
| HELM Capabilities mean | HELM Capabilities | – | 79.9% |
| Vals Finance Index | Vals AI | 52.9%† | – |
| Vals Router Index | Vals AI | 53.0%† | – |
| Vals Legal Index | Vals AI | 69.0%† | – |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | – | 41.7% |
| MASK | Scale AI SEAL | – | 42.6% |
| PropensityBench | Scale AI SEAL | – | 52.9% |
| Metadata | |||
| AA Openness Index | Artificial Analysis | 44.4 | 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
- Gemini 3 Pro: 1 score revision.
- DeepSeek V4 Pro (high): BenchLeader Index 59.1 → 56.0.
- DeepSeek V4 Pro (high): 17 new results (AA Intelligence Index v5.0.1, Humanity's Last Exam (AA), HLE-Diamond (AA)…); input price moved from $0.66/M to $1.32/M (100% up); output price moved from $1.98/M to $3.96/M (100% up).
- DeepSeek V4 Pro (high): BenchLeader Index 57.5 → 59.1.
- Gemini 3 Pro: 1 new result (Arena Search).
- DeepSeek V4 Pro (high): 1 score revision.
- DeepSeek V4 Pro (high): 1 score revision.
- Gemini 3 Pro: 1 score revision.
- DeepSeek V4 Pro (high): 16 new results (GPQA Diamond, OTIS Mock AIME, SciCode…); input price moved from $0.44/M to $0.66/M (50% up); output price moved from $0.87/M to $1.98/M (128% up).
- Gemini 3 Pro: 1 score revision.
Related comparisons
- DeepSeek V4 Pro vs Claude Fable 5.2
- DeepSeek V4 Pro vs Claude Opus 5.5
- Gemini 3 Pro vs Gemini 4 Carbon
- Gemini 3 Pro vs Claude Opus 5.5
- Gemini 3 Pro vs GPT-6 Astra
- DeepSeek V4 Flash vs Gemini 3 Pro
- DeepSeek V4 Pro vs Claude Fable 5.1
- DeepSeek V4 Pro vs GPT-6.1 Sol
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
DeepSeek V4 Pro vs Gemini 3 Pro: questions
- Is DeepSeek V4 Pro better than Gemini 3 Pro?
- DeepSeek V4 Pro (high) and Gemini 3 Pro are neck and neck on quality (56.0 vs 60.0, within the ±5.6 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.
- DeepSeek V4 Pro vs Gemini 3 Pro: which is cheaper, and will that change?
- DeepSeek V4 Pro is cheaper today: $0.990 against $4.50 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is DeepSeek V4 Pro or Gemini 3 Pro better for coding?
- Gemini 3 Pro scores higher in coding (58 vs 52 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1581 for DeepSeek V4 Pro against 1441 for Gemini 3 Pro.
- Do independent benchmarks agree on DeepSeek V4 Pro vs Gemini 3 Pro?
- Across the 15 benchmarks both have, DeepSeek V4 Pro leads 4, Gemini 3 Pro leads 11. Every overall index that rates both puts Gemini 3 Pro ahead.
- When was DeepSeek V4 Pro released?
- DeepSeek V4 Pro was released on 13 Aug 2026.