DeepSeek V4 Pro 0813 vs Gemini 2.5 Pro
Updated October 21, 2026. 49 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Pick DeepSeek V4 Pro 0813 for coding: it scores 54.6 on the Vibe Coding Index against 37.0 for Gemini 2.5 Pro, though on 1 benchmark against 2.
Pick Gemini 2.5 Pro when what matters is LMArena Text: math or LMArena Text: mathematical work, where it leads.
- Shared results
- 49
- DeepSeek V4 Pro 0813 only
- 29
- Gemini 2.5 Pro only
- 38
- Like-for-like categories
- 7 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, four public coding benchmarks
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 scores 54.6 against 37.0.
1 and 2 benchmarks behind the scores
Intelligence
AA Intelligence Index
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 leads, 36.0 against 16.1.
Same board, both measured
Agentic work
AA Agentic Index
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 leads, 41.3 against 1.6.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only DeepSeek V4 Pro 0813 has a GPQA Diamond result, 92.4%.
Directional only
Knowledge
MMLU Pro
No call
Only DeepSeek V4 Pro 0813 has a MMLU Pro result, 87.0%.
Directional only
Web design
Design Arena: websites
No call
Only Gemini 2.5 Pro has a Design Arena: websites result, 1167.
Directional only
General
Vals Index
No call
Only DeepSeek V4 Pro 0813 has a Vals Index result, 53.0%.
Directional only
Long documents
Listed context window
No call
Both take 1.05M tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 $0.0016, Gemini 2.5 Pro $0.0063 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 $0.10, Gemini 2.5 Pro $0.64.
Published attempt cost and pass rate
At a glance
- Coding index
- DeepSeek V4 Pro 0813 54.6 from 1 benchmark, Gemini 2.5 Pro 37.0 from 2 benchmarks.
- Widest category
- Agentic work, 39.7 points apart on average over 1 shared benchmark.
- Widest single gap
- Vibe Code Bench (Vals AI): DeepSeek V4 Pro 0813 82.3% against Gemini 2.5 Pro 0.4%, 81.9 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V4 Pro 0813 $1.98 against Gemini 2.5 Pro $10.00 per million output tokens, 5.1x apart.
- Context window
- Both take 1.05M tokens.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | DeepSeek V4 Pro 0813 | Bars | Gemini 2.5 Pro | Ties |
|---|---|---|---|---|
| Coding | 9 | 0 | ||
| Intelligence | 2 | 0 | ||
| Agentic work | 1 | 0 | ||
| Reasoning | 8 | 0 | ||
| Math | 3 | 2 | ||
| Knowledge | 2 | 4 | ||
| General | 8 | 10 | ||
| All shared boards | 33 | 16 |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index90
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 2.5 Pro | Leader |
|---|---|---|---|
| SWE-bench Verified | Not measured | 53.6% | Only Gemini 2.5 Pro measured |
| LMArena WebDev | 1582 | 1226 | DeepSeek V4 Pro 0813 |
| AA Coding Index | 68.8 | 33.3 | DeepSeek V4 Pro 0813 |
| LiveCodeBench (Vals AI) | 87.5% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| SWE-bench Verified (Vals AI) | 96.4% | 54.4% | DeepSeek V4 Pro 0813 |
| Terminal-Bench 4 (Vals AI) | 14.1% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Vibe Code Bench (Vals AI) | 82.3% | 0.4% | DeepSeek V4 Pro 0813 |
| Code migration (Vals AI) | 41.5% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| ProgramBench (Vals AI) | 0.0% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| IOI (Vals AI) | 51.6% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Terminal-Bench 2.1 (Vals AI) | 54.7% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Vibe Code Bench 1-100 (Vals AI) | 17.5% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| SkillsBench (Vals AI) | 53.8% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| CyberBench (Vals AI) | 72.9% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 80.1% | Only Gemini 2.5 Pro measured |
| SciCode | 51.0% | 46.3% | DeepSeek V4 Pro 0813 |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 26.5% | Only Gemini 2.5 Pro measured |
| Terminal-Bench | Not measured | 32.6% | Only Gemini 2.5 Pro measured |
| FrontierCode | 28.5% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| GSO-Bench | Not measured | 3.9% | Only Gemini 2.5 Pro measured |
| WeirdML | 66.2% | 54.0% | DeepSeek V4 Pro 0813 |
| ALE-Bench | 1403.2 | Not measured | Only DeepSeek V4 Pro 0813 measured |
| LMArena Text: coding | 1469 | 1452 | DeepSeek V4 Pro 0813 |
| LMArena Text: software and IT | 1462 | 1459 | DeepSeek V4 Pro 0813 |
| LMArena WebDev: React | 1588 | Not measured | Only DeepSeek V4 Pro 0813 measured |
| LMArena WebDev: HTML | 1566 | 1235 | DeepSeek V4 Pro 0813 |
| LMArena WebDev: from an image | Not measured | 1285 | Only Gemini 2.5 Pro measured |
| Neither model has a result on the other 11 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 2.5 Pro | Leader |
|---|---|---|---|
| AA Intelligence Index | 36.0 | 16.1 | DeepSeek V4 Pro 0813 |
| Epoch Capabilities Index | 155.6 | 145.1 | DeepSeek V4 Pro 0813 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 2.5 Pro | Leader |
|---|---|---|---|
| AA Agentic Index | 41.3 | 1.6 | DeepSeek V4 Pro 0813 |
| SREBench (Vals AI) | 1.9% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Finance Agent (Vals AI) | 50.4% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Tax Agent Bench (Vals AI) | 26.8% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Legal Research Bench (Vals AI) | 40.9% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Harvey's Legal Agent Benchmark (Vals AI) | 10.8% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Excel Modeling Benchmark (Vals AI) | 52.8% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Terminal-Bench-Science (Vals AI) | 4.3% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| τ²-Bench Telecom | Not measured | 54.1% | Only Gemini 2.5 Pro measured |
| APEX-Agents | 47.3% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| GDPval | Not measured | 23.3% | Only Gemini 2.5 Pro measured |
| DeepResearch Bench | Not measured | 41.5% | Only Gemini 2.5 Pro measured |
| Vending-Bench 2 | Not measured | $574 | Only Gemini 2.5 Pro measured |
| Neither model has a result on the other 26 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index80
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 2.5 Pro | Leader |
|---|---|---|---|
| GPQA Diamond | 92.4% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| GPQA Diamond (Artificial Analysis run) | 92.8% | 84.4% | DeepSeek V4 Pro 0813 |
| Humanity's Last Exam (Artificial Analysis run) | 41.0% | 22.5% | DeepSeek V4 Pro 0813 |
| GPQA Diamond (Epoch AI run) | 91.7% | 85.3% | DeepSeek V4 Pro 0813 |
| ARC-AGI | 90.5% | 41.0% | DeepSeek V4 Pro 0813 |
| ARC-AGI-2 | 61.3% | 4.9% | DeepSeek V4 Pro 0813 |
| Chess puzzles | 46.0% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Mystery game puzzles | 43.0% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| LMArena Text: hard prompts | 1458 | 1455 | DeepSeek V4 Pro 0813 |
| LMArena Text: expert | 1459 | 1452 | DeepSeek V4 Pro 0813 |
| LMArena Text: hard prompts in English | 1463 | 1453 | DeepSeek V4 Pro 0813 |
| LMArena Vision: diagrams | Not measured | 1274 | Only Gemini 2.5 Pro measured |
| LMArena Vision: homework | Not measured | 1281 | Only Gemini 2.5 Pro measured |
| Neither model has a result on the other 8 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index32
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 2.5 Pro | Leader |
|---|---|---|---|
| AA Math Index | Not measured | 87.7 | Only Gemini 2.5 Pro measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 87.7% | Only Gemini 2.5 Pro measured |
| OTIS Mock AIME | 98.6% | 84.7% | DeepSeek V4 Pro 0813 |
| FrontierMath, tiers 1 to 3 | 64.6% | 24.6% | DeepSeek V4 Pro 0813 |
| FrontierMath, tier 4 | 26.8% | 0.0% | DeepSeek V4 Pro 0813 |
| ProofBench | 50.0% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| LMArena Text: math | 1441 | 1450 | Gemini 2.5 Pro |
| LMArena Text: mathematical work | 1443 | 1454 | Gemini 2.5 Pro |
| Neither model has a result on the other 5 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index24
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 2.5 Pro | Leader |
|---|---|---|---|
| MMLU Pro | 87.0% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| LegalBench (Vals AI) | 82.4% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| CaseLaw (Vals AI) | Not measured | 63.9% | Only Gemini 2.5 Pro measured |
| CorpFin (Vals AI) | 65.4% | 60.8% | DeepSeek V4 Pro 0813 |
| TaxEval (Vals AI) | 73.1% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| MortgageTax (Vals AI) | Not measured | 68.9% | Only Gemini 2.5 Pro measured |
| MedCode (Vals AI) | 42.5% | 50.6% | Gemini 2.5 Pro |
| MMLU Pro (Artificial Analysis run) | Not measured | 86.2% | Only Gemini 2.5 Pro measured |
| SimpleQA Verified | 52.9% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| LMArena Text: science | 1457 | 1480 | Gemini 2.5 Pro |
| LMArena Text: medicine and healthcare | 1446 | 1476 | Gemini 2.5 Pro |
| LMArena Text: legal and government | 1459 | 1480 | Gemini 2.5 Pro |
| LMArena Text: business and finance | 1442 | 1439 | DeepSeek V4 Pro 0813 |
| Neither model has a result on the other 4 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 2.5 Pro | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1167 | Only Gemini 2.5 Pro measured |
| Design Arena: UI components | Not measured | 1139 | Only Gemini 2.5 Pro measured |
| Design Arena: game dev | Not measured | 1112 | Only Gemini 2.5 Pro measured |
| Design Arena: data viz | Not measured | 1222 | Only Gemini 2.5 Pro measured |
| Design Arena: 3D | Not measured | 1080 | Only Gemini 2.5 Pro measured |
| Design Arena: all code categories | Not measured | 1152 | Only Gemini 2.5 Pro measured |
| Neither model has a result on the other 3 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index810
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 2.5 Pro | Leader |
|---|---|---|---|
| Vals Index | 53.0% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| SAGE (Vals AI) | Not measured | 41.9% | Only Gemini 2.5 Pro measured |
| MedScribe (Vals AI) | 80.2% | 73.5% | DeepSeek V4 Pro 0813 |
| IFBench | Not measured | 48.7% | Only Gemini 2.5 Pro measured |
| Long context reasoning (AA-LCR) | 80.3% | 69.0% | DeepSeek V4 Pro 0813 |
| DTBench | 93.9% | 82.4% | DeepSeek V4 Pro 0813 |
| LMCA | 45.5% | 34.8% | DeepSeek V4 Pro 0813 |
| Creative writing (Lech Mazur) | Not measured | 83.8% | Only Gemini 2.5 Pro measured |
| LMArena Text | 1446 | 1458 | Gemini 2.5 Pro |
| LMArena Text: instruction following | 1448 | 1437 | DeepSeek V4 Pro 0813 |
| LMArena Text: creative writing | 1439 | 1454 | Gemini 2.5 Pro |
| LMArena Vision | Not measured | 1263 | Only Gemini 2.5 Pro measured |
| LMArena Text: English | 1457 | 1458 | Gemini 2.5 Pro |
| LMArena Text: non-English | 1431 | 1452 | Gemini 2.5 Pro |
| LMArena Text: Chinese | 1481 | 1507 | Gemini 2.5 Pro |
| LMArena Text: German | Not measured | 1487 | Only Gemini 2.5 Pro measured |
| LMArena Text: Japanese | Not measured | 1461 | Only Gemini 2.5 Pro measured |
| LMArena Text: Korean | 1449 | 1434 | DeepSeek V4 Pro 0813 |
| LMArena Text: French | 1446 | 1473 | Gemini 2.5 Pro |
| LMArena Text: Spanish | 1435 | 1474 | Gemini 2.5 Pro |
| LMArena Text: Russian | 1448 | 1461 | Gemini 2.5 Pro |
| LMArena Text: Polish | Not measured | 1479 | Only Gemini 2.5 Pro measured |
| LMArena Text: multi-turn | 1454 | 1453 | DeepSeek V4 Pro 0813 |
| LMArena Text: longer queries | 1458 | 1449 | DeepSeek V4 Pro 0813 |
| LMArena Text: writing and language | 1443 | 1454 | Gemini 2.5 Pro |
| LMArena Text: entertainment and media | 1424 | 1441 | Gemini 2.5 Pro |
| LMArena Vision: English | Not measured | 1261 | Only Gemini 2.5 Pro measured |
| LMArena Vision: Chinese | Not measured | 1300 | Only Gemini 2.5 Pro measured |
| LMArena Vision: reading text | Not measured | 1270 | Only Gemini 2.5 Pro measured |
| LMArena Vision: humor | Not measured | 1262 | Only Gemini 2.5 Pro measured |
| LMArena Vision: captioning | Not measured | 1304 | Only Gemini 2.5 Pro measured |
| LMArena Vision: recognizing things | Not measured | 1283 | Only Gemini 2.5 Pro measured |
| LMArena Vision: creative writing | Not measured | 1265 | Only Gemini 2.5 Pro measured |
| LMArena Document | Not measured | 1421 | Only Gemini 2.5 Pro measured |
| Neither model has a result on the other 12 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | DeepSeek V4 Pro 0813 | Gemini 2.5 Pro | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
DeepSeek V4 Pro 0813 costs less
Repository review
50K input + 3K output tokens
DeepSeek V4 Pro 0813 costs less
Agent loop
200K input + 10K output tokens
DeepSeek V4 Pro 0813 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | DeepSeek V4 Pro 0813 | Google Gemini 2.5 Pro |
|---|---|---|
| Maker | DeepSeek | |
| Weights | Open weights | No public weights listed |
| Released | August 12, 2026 | June 17, 2025 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $0.66 | $1.25 |
| Output $/Mtok | $1.98 | $10.00 |
| Coding index | 54.6 | 37.0 |
| Independent measurements behind it | 1 | 2 |
| Cost per solved task | $0.10 | $0.64 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V4 Pro 0813 better than Gemini 2.5 Pro for coding?
Pick DeepSeek V4 Pro 0813 for coding: it scores 54.6 on the Vibe Coding Index against 37.0 for Gemini 2.5 Pro, though on 1 benchmark against 2. Pick Gemini 2.5 Pro when what matters is LMArena Text: math or LMArena Text: mathematical work, where it leads.
Which is cheaper, DeepSeek V4 Pro 0813 or Gemini 2.5 Pro?
DeepSeek V4 Pro 0813 costs $0.66 per million input tokens and $1.98 per million output tokens; Gemini 2.5 Pro costs $1.25 and $10.00. Per solved SWE-bench Verified task, DeepSeek V4 Pro 0813 costs $0.10 and Gemini 2.5 Pro $0.64.
Which has the bigger context window, DeepSeek V4 Pro 0813 or Gemini 2.5 Pro?
Neither: both take 1.05M tokens.
How many benchmarks is the DeepSeek V4 Pro 0813 vs Gemini 2.5 Pro comparison based on?
49 benchmarks have a result for both models as of October 21, 2026. DeepSeek V4 Pro 0813's coding index rests on 1 independent benchmark and Gemini 2.5 Pro's on 2. Where only one model has a result, the page says so instead of naming a winner.