DeepSeek V4 Flash Vision Exp vs Gemini 3.7 Flash
Updated October 21, 2026. 13 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Pick Gemini 3.7 Flash for coding: it scores 69.9 on the Vibe Coding Index against 37.6 for DeepSeek V4 Flash Vision Exp.
Pick DeepSeek V4 Flash Vision Exp when output tokens cost 5.8x less, or when what matters is LiveBench browser or LiveBench long agentic, where it leads.
- Shared results
- 13
- DeepSeek V4 Flash Vision Exp only
- 1
- Gemini 3.7 Flash only
- 99
- Like-for-like categories
- 6 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
Gemini 3.7 Flash
Gemini 3.7 Flash scores 69.9 against 37.6.
1 and 2 benchmarks behind the scores
Intelligence
AA Intelligence Index
Gemini 3.7 Flash
Gemini 3.7 Flash leads, 39.6 against 34.8.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only DeepSeek V4 Flash Vision Exp has a AA Agentic Index result, 47.5.
Directional only
Reasoning
GPQA Diamond
No call
Only Gemini 3.7 Flash has a GPQA Diamond result, 93.9%.
Directional only
Knowledge
MMLU Pro
No call
Only Gemini 3.7 Flash has a MMLU Pro result, 90.1%.
Directional only
Web design
Design Arena: websites
No call
Only Gemini 3.7 Flash has a Design Arena: websites result, 1304.
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 Flash Vision Exp
DeepSeek V4 Flash Vision Exp $0.0005, Gemini 3.7 Flash $0.0026 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Gemini 3.7 Flash has the number, $1.78.
Directional only
At a glance
- Coding index
- DeepSeek V4 Flash Vision Exp 37.6 from 1 benchmark, Gemini 3.7 Flash 69.9 from 2 benchmarks.
- Widest category
- Coding, 19.8 points apart on average over 3 shared benchmarks.
- Widest single gap
- LiveBench coding: DeepSeek V4 Flash Vision Exp 37.6% against Gemini 3.7 Flash 80.5%, 42.9 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V4 Flash Vision Exp $0.6468 against Gemini 3.7 Flash $3.75 per million output tokens, 5.8x 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 Flash Vision Exp | Bars | Gemini 3.7 Flash | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 2 | 1 | ||
| Reasoning | 0 | 3 | ||
| Web design | 0 | 1 | ||
| General | 1 | 1 | ||
| All shared boards | 3 | 10 |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index03
| Benchmark | DeepSeek V4 Flash Vision Exp | Gemini 3.7 Flash | Leader |
|---|---|---|---|
| LiveBench coding | 37.6% | 80.5% | Gemini 3.7 Flash |
| LMArena WebDev | Not measured | 1591 | Only Gemini 3.7 Flash measured |
| AA Coding Index | 65.0 | 71.5 | Gemini 3.7 Flash |
| LiveCodeBench (Vals AI) | Not measured | 88.7% | Only Gemini 3.7 Flash measured |
| SWE-bench Verified (Vals AI) | Not measured | 80.8% | Only Gemini 3.7 Flash measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 12.1% | Only Gemini 3.7 Flash measured |
| Vibe Code Bench (Vals AI) | Not measured | 70.4% | Only Gemini 3.7 Flash measured |
| Code migration (Vals AI) | Not measured | 34.8% | Only Gemini 3.7 Flash measured |
| ProgramBench (Vals AI) | Not measured | 0.0% | Only Gemini 3.7 Flash measured |
| IOI (Vals AI) | Not measured | 67.8% | Only Gemini 3.7 Flash measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 77.5% | Only Gemini 3.7 Flash measured |
| SkillsBench (Vals AI) | Not measured | 65.9% | Only Gemini 3.7 Flash measured |
| CyberBench (Vals AI) | Not measured | 43.8% | Only Gemini 3.7 Flash measured |
| SciCode | 49.7% | 59.8% | Gemini 3.7 Flash |
| FrontierCode | Not measured | 43.6% | Only Gemini 3.7 Flash measured |
| FrontierSWE | Not measured | 20.3% | Only Gemini 3.7 Flash measured |
| ALE-Bench | Not measured | 904.3 | Only Gemini 3.7 Flash measured |
| LMArena Text: coding | Not measured | 1499 | Only Gemini 3.7 Flash measured |
| LMArena Text: software and IT | Not measured | 1498 | Only Gemini 3.7 Flash measured |
| LMArena WebDev: React | Not measured | 1587 | Only Gemini 3.7 Flash measured |
| LMArena WebDev: HTML | Not measured | 1593 | Only Gemini 3.7 Flash measured |
| LMArena WebDev: from an image | Not measured | 1599 | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 16 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V4 Flash Vision Exp | Gemini 3.7 Flash | Leader |
|---|---|---|---|
| AA Intelligence Index | 34.8 | 39.6 | Gemini 3.7 Flash |
| Epoch Capabilities Index | Not measured | 158.0 | Only Gemini 3.7 Flash measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index21
| Benchmark | DeepSeek V4 Flash Vision Exp | Gemini 3.7 Flash | Leader |
|---|---|---|---|
| AA Agentic Index | 47.5 | Not measured | Only DeepSeek V4 Flash Vision Exp measured |
| Design Arena: agents building web apps | Not measured | 1224 | Only Gemini 3.7 Flash measured |
| Design Arena: agents building full-stack apps | Not measured | 1208 | Only Gemini 3.7 Flash measured |
| Design Arena: agents building mobile apps | Not measured | 1230 | Only Gemini 3.7 Flash measured |
| Design Arena: agents building Android apps | Not measured | 1247 | Only Gemini 3.7 Flash measured |
| Design Arena: agents building Godot games | Not measured | 1264 | Only Gemini 3.7 Flash measured |
| Design Arena: agents building games | Not measured | 1235 | Only Gemini 3.7 Flash measured |
| Design Arena: spreadsheets | Not measured | 1028 | Only Gemini 3.7 Flash measured |
| LiveBench research | 69.2% | 76.5% | Gemini 3.7 Flash |
| LiveBench browser | 64.0% | 56.0% | DeepSeek V4 Flash Vision Exp |
| LiveBench long agentic | 39.0% | 28.3% | DeepSeek V4 Flash Vision Exp |
| SREBench (Vals AI) | Not measured | 4.6% | Only Gemini 3.7 Flash measured |
| Finance Agent (Vals AI) | Not measured | 59.0% | Only Gemini 3.7 Flash measured |
| Tax Agent Bench (Vals AI) | Not measured | 22.8% | Only Gemini 3.7 Flash measured |
| Legal Research Bench (Vals AI) | Not measured | 34.6% | Only Gemini 3.7 Flash measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 71.3% | Only Gemini 3.7 Flash measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 7.1% | Only Gemini 3.7 Flash measured |
| APEX-Agents | Not measured | 67.8% | Only Gemini 3.7 Flash measured |
| Remote Labor Index | Not measured | 5.0% | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 20 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index03
| Benchmark | DeepSeek V4 Flash Vision Exp | Gemini 3.7 Flash | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 93.9% | Only Gemini 3.7 Flash measured |
| LiveBench reasoning | 85.4% | 87.8% | Gemini 3.7 Flash |
| MMMU Pro (Vals AI) | Not measured | 89.0% | Only Gemini 3.7 Flash measured |
| GPQA Diamond (Artificial Analysis run) | 91.3% | 94.5% | Gemini 3.7 Flash |
| Humanity's Last Exam (Artificial Analysis run) | 34.5% | 47.9% | Gemini 3.7 Flash |
| GPQA Diamond (Epoch AI run) | Not measured | 94.8% | Only Gemini 3.7 Flash measured |
| ARC-AGI | Not measured | 95.5% | Only Gemini 3.7 Flash measured |
| ARC-AGI-2 | Not measured | 84.6% | Only Gemini 3.7 Flash measured |
| Chess puzzles | Not measured | 66.0% | Only Gemini 3.7 Flash measured |
| Mystery game puzzles | Not measured | 37.0% | Only Gemini 3.7 Flash measured |
| LMArena Text: hard prompts | Not measured | 1495 | Only Gemini 3.7 Flash measured |
| LMArena Text: expert | Not measured | 1512 | Only Gemini 3.7 Flash measured |
| LMArena Text: hard prompts in English | Not measured | 1487 | Only Gemini 3.7 Flash measured |
| LMArena Vision: diagrams | Not measured | 1329 | Only Gemini 3.7 Flash measured |
| LMArena Vision: homework | Not measured | 1320 | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 6 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | DeepSeek V4 Flash Vision Exp | Gemini 3.7 Flash | Leader |
|---|---|---|---|
| OTIS Mock AIME | Not measured | 97.2% | Only Gemini 3.7 Flash measured |
| FrontierMath, tiers 1 to 3 | Not measured | 71.6% | Only Gemini 3.7 Flash measured |
| FrontierMath, tier 4 | Not measured | 36.6% | Only Gemini 3.7 Flash measured |
| ProofBench | Not measured | 58.0% | Only Gemini 3.7 Flash measured |
| LMArena Text: math | Not measured | 1511 | Only Gemini 3.7 Flash measured |
| LMArena Text: mathematical work | Not measured | 1502 | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index10 boards, none shared
| Benchmark | DeepSeek V4 Flash Vision Exp | Gemini 3.7 Flash | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 90.1% | Only Gemini 3.7 Flash measured |
| LegalBench (Vals AI) | Not measured | 87.3% | Only Gemini 3.7 Flash measured |
| TaxEval (Vals AI) | Not measured | 74.7% | Only Gemini 3.7 Flash measured |
| MortgageTax (Vals AI) | Not measured | 66.7% | Only Gemini 3.7 Flash measured |
| MedCode (Vals AI) | Not measured | 53.4% | Only Gemini 3.7 Flash measured |
| SimpleQA Verified | Not measured | 69.2% | Only Gemini 3.7 Flash measured |
| LMArena Text: science | Not measured | 1511 | Only Gemini 3.7 Flash measured |
| LMArena Text: medicine and healthcare | Not measured | 1494 | Only Gemini 3.7 Flash measured |
| LMArena Text: legal and government | Not measured | 1496 | Only Gemini 3.7 Flash measured |
| LMArena Text: business and finance | Not measured | 1479 | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 7 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V4 Flash Vision Exp | Gemini 3.7 Flash | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1304 | Only Gemini 3.7 Flash measured |
| Design Arena: UI components | Not measured | 1299 | Only Gemini 3.7 Flash measured |
| Design Arena: game dev | Not measured | 1318 | Only Gemini 3.7 Flash measured |
| Design Arena: data viz | Not measured | 1315 | Only Gemini 3.7 Flash measured |
| Design Arena: 3D | Not measured | 1308 | Only Gemini 3.7 Flash measured |
| Design Arena: all code categories | Not measured | 1308 | Only Gemini 3.7 Flash measured |
| Design Arena: SVG | Not measured | 1283 | Only Gemini 3.7 Flash measured |
| Design Arena: ASCII art | Not measured | 1276 | Only Gemini 3.7 Flash measured |
| LiveBench design | 70.7% | 73.2% | Gemini 3.7 Flash |
GeneralLike-for-likePublic boards, shown as published, kept out of the index11
| Benchmark | DeepSeek V4 Flash Vision Exp | Gemini 3.7 Flash | Leader |
|---|---|---|---|
| LiveBench data analysis | 79.5% | 68.0% | DeepSeek V4 Flash Vision Exp |
| SAGE (Vals AI) | Not measured | 49.2% | Only Gemini 3.7 Flash measured |
| MedScribe (Vals AI) | Not measured | 83.9% | Only Gemini 3.7 Flash measured |
| Long context reasoning (AA-LCR) | 81.3% | 83.0% | Gemini 3.7 Flash |
| DTBench | Not measured | 96.8% | Only Gemini 3.7 Flash measured |
| LMCA | Not measured | 50.4% | Only Gemini 3.7 Flash measured |
| Furniture assembly | Not measured | 26.7% | Only Gemini 3.7 Flash measured |
| LMArena Text | Not measured | 1489 | Only Gemini 3.7 Flash measured |
| LMArena Text: instruction following | Not measured | 1484 | Only Gemini 3.7 Flash measured |
| LMArena Text: creative writing | Not measured | 1488 | Only Gemini 3.7 Flash measured |
| LMArena Vision | Not measured | 1314 | Only Gemini 3.7 Flash measured |
| LMArena Text: English | Not measured | 1483 | Only Gemini 3.7 Flash measured |
| LMArena Text: non-English | Not measured | 1487 | Only Gemini 3.7 Flash measured |
| LMArena Text: Chinese | Not measured | 1552 | Only Gemini 3.7 Flash measured |
| LMArena Text: German | Not measured | 1506 | Only Gemini 3.7 Flash measured |
| LMArena Text: Japanese | Not measured | 1511 | Only Gemini 3.7 Flash measured |
| LMArena Text: Korean | Not measured | 1481 | Only Gemini 3.7 Flash measured |
| LMArena Text: French | Not measured | 1500 | Only Gemini 3.7 Flash measured |
| LMArena Text: Spanish | Not measured | 1495 | Only Gemini 3.7 Flash measured |
| LMArena Text: Russian | Not measured | 1515 | Only Gemini 3.7 Flash measured |
| LMArena Text: Polish | Not measured | 1503 | Only Gemini 3.7 Flash measured |
| LMArena Text: multi-turn | Not measured | 1485 | Only Gemini 3.7 Flash measured |
| LMArena Text: longer queries | Not measured | 1492 | Only Gemini 3.7 Flash measured |
| LMArena Text: writing and language | Not measured | 1491 | Only Gemini 3.7 Flash measured |
| LMArena Text: entertainment and media | Not measured | 1477 | Only Gemini 3.7 Flash measured |
| LMArena Vision: English | Not measured | 1309 | Only Gemini 3.7 Flash measured |
| LMArena Vision: Chinese | Not measured | 1394 | Only Gemini 3.7 Flash measured |
| LMArena Vision: reading text | Not measured | 1322 | Only Gemini 3.7 Flash measured |
| LMArena Vision: humor | Not measured | 1299 | Only Gemini 3.7 Flash measured |
| LMArena Vision: creative writing | Not measured | 1312 | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 16 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | DeepSeek V4 Flash Vision Exp | Gemini 3.7 Flash | 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 Flash Vision Exp costs less
Repository review
50K input + 3K output tokens
DeepSeek V4 Flash Vision Exp costs less
Agent loop
200K input + 10K output tokens
DeepSeek V4 Flash Vision Exp 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 Flash Vision Exp | Google Gemini 3.7 Flash |
|---|---|---|
| Maker | DeepSeek | |
| Weights | Open weights | No public weights listed |
| Released | August 21, 2026 | August 13, 2026 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $0.2156 | $0.75 |
| Output $/Mtok | $0.6468 | $3.75 |
| Coding index | 37.6 | 69.9 |
| Independent measurements behind it | 1 | 2 |
| Cost per solved task | Not computable | $1.78 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V4 Flash Vision Exp better than Gemini 3.7 Flash for coding?
Pick Gemini 3.7 Flash for coding: it scores 69.9 on the Vibe Coding Index against 37.6 for DeepSeek V4 Flash Vision Exp. Pick DeepSeek V4 Flash Vision Exp when output tokens cost 5.8x less, or when what matters is LiveBench browser or LiveBench long agentic, where it leads.
Which is cheaper, DeepSeek V4 Flash Vision Exp or Gemini 3.7 Flash?
DeepSeek V4 Flash Vision Exp costs $0.2156 per million input tokens and $0.6468 per million output tokens; Gemini 3.7 Flash costs $0.75 and $3.75. Cost per solved task is known for Gemini 3.7 Flash only.
Which has the bigger context window, DeepSeek V4 Flash Vision Exp or Gemini 3.7 Flash?
Neither: both take 1.05M tokens.
How many benchmarks is the DeepSeek V4 Flash Vision Exp vs Gemini 3.7 Flash comparison based on?
13 benchmarks have a result for both models as of October 21, 2026. DeepSeek V4 Flash Vision Exp's coding index rests on 1 independent benchmark and Gemini 3.7 Flash's on 2. Where only one model has a result, the page says so instead of naming a winner.