Gemini 2.5 Flash vs Gemma 3 4B
Gemini 2.5 Flash leads Gemma 3 4B on the Artificial Analysis Intelligence Index, 15.5 to 4.8, and Gemma 3 4B costs 33.3x less per output token ($0.075 vs $2.50 per million). Updated October 21, 2026; 36 benchmarks have a result for both.
Gemini 2.5 Flash or Gemma 3 4B: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick Gemini 2.5 Flash for a larger context window (1.05M against 131K tokens).
- Shared results
- 36
- Gemini 2.5 Flash only
- 33
- Gemma 3 4B only
- 3
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
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, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
Gemini 2.5 Flash
Gemini 2.5 Flash leads, 15.5 against 4.8.
Same board, both measured
Web design
Design Arena: websites
No call
Only Gemini 2.5 Flash has a Design Arena: websites result, 1074.
Directional only
Long documents
Listed context window
Gemini 2.5 Flash
Gemini 2.5 Flash takes 1.05M tokens against 131K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Gemma 3 4B
Gemini 2.5 Flash $0.0015, Gemma 3 4B $0.0001 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Gemini 2.5 Flash has the number, $0.466.
Directional only
At a glance
- Coding index
- Gemini 2.5 Flash not benchmarked, Gemma 3 4B not benchmarked.
- Widest category
- Math, 44.4 points apart on average over 4 shared benchmarks.
- Widest single gap
- AIME 2025 (Artificial Analysis run): Gemini 2.5 Flash 73.3% against Gemma 3 4B 12.7%, 60.7 points apart on a 0 to 100 scale.
- Output price
- Gemini 2.5 Flash $2.50 against Gemma 3 4B $0.075 per million output tokens, 33.3x apart.
- Context window
- Gemini 2.5 Flash takes 1.05M tokens, Gemma 3 4B 131K.
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 | Gemini 2.5 Flash | Bars | Gemma 3 4B | Ties |
|---|---|---|---|---|
| Coding | 4 | 0 | ||
| Intelligence | 2 | 0 | ||
| Agentic work | 1 | 0 | ||
| Reasoning | 5 | 0 | ||
| Math | 4 | 0 | ||
| Knowledge | 5 | 0 | ||
| General | 15 | 0 | ||
| All shared boards | 36 | 0 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index40
| Benchmark | Gemini 2.5 Flash | Gemma 3 4B | Leader |
|---|---|---|---|
| SWE-bench Verified | 28.7% | Not measured | Only Gemini 2.5 Flash measured |
| AA Coding Index | Not measured | 2.7 | Only Gemma 3 4B measured |
| LiveCodeBench (Artificial Analysis run) | 69.5% | 11.2% | Gemini 2.5 Flash |
| Terminal-Bench Hard (Artificial Analysis run) | 13.6% | 0.8% | Gemini 2.5 Flash |
| Terminal-Bench | 17.1% | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Text: coding | 1423 | 1230 | Gemini 2.5 Flash |
| LMArena Text: software and IT | 1425 | 1269 | Gemini 2.5 Flash |
| Neither model has a result on the other 31 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | Gemini 2.5 Flash | Gemma 3 4B | Leader |
|---|---|---|---|
| AA Intelligence Index | 15.5 | 4.8 | Gemini 2.5 Flash |
| Epoch Capabilities Index | 142.9 | 115.9 | Gemini 2.5 Flash |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | Gemini 2.5 Flash | Gemma 3 4B | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 31.6% | 5.0% | Gemini 2.5 Flash |
| BALROG | 33.5% | Not measured | Only Gemini 2.5 Flash measured |
| Vending-Bench 2 | $549 | Not measured | Only Gemini 2.5 Flash measured |
| Neither model has a result on the other 36 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | Gemini 2.5 Flash | Gemma 3 4B | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 79.0% | 29.1% | Gemini 2.5 Flash |
| Humanity's Last Exam (Artificial Analysis run) | 12.1% | 5.3% | Gemini 2.5 Flash |
| GPQA Diamond (Epoch AI run) | Not measured | 23.2% | Only Gemma 3 4B measured |
| ARC-AGI | 33.3% | Not measured | Only Gemini 2.5 Flash measured |
| ARC-AGI-2 | 2.5% | Not measured | Only Gemini 2.5 Flash measured |
| SimpleBench | 41.2% | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Text: hard prompts | 1422 | 1253 | Gemini 2.5 Flash |
| LMArena Text: expert | 1426 | 1223 | Gemini 2.5 Flash |
| LMArena Text: hard prompts in English | 1420 | 1253 | Gemini 2.5 Flash |
| LMArena Vision: diagrams | 1245 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Vision: homework | 1244 | Not measured | Only Gemini 2.5 Flash measured |
| Neither model has a result on the other 10 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index40
| Benchmark | Gemini 2.5 Flash | Gemma 3 4B | Leader |
|---|---|---|---|
| AA Math Index | 73.3 | 12.7 | Gemini 2.5 Flash |
| AIME 2025 (Artificial Analysis run) | 73.3% | 12.7% | Gemini 2.5 Flash |
| OTIS Mock AIME | Not measured | 7.5% | Only Gemma 3 4B measured |
| LMArena Text: math | 1411 | 1239 | Gemini 2.5 Flash |
| LMArena Text: mathematical work | 1422 | 1256 | Gemini 2.5 Flash |
| Neither model has a result on the other 8 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | Gemini 2.5 Flash | Gemma 3 4B | Leader |
|---|---|---|---|
| MedCode (Vals AI) | 40.4% | Not measured | Only Gemini 2.5 Flash measured |
| MMLU Pro (Artificial Analysis run) | 83.2% | 41.7% | Gemini 2.5 Flash |
| LMArena Text: science | 1433 | 1306 | Gemini 2.5 Flash |
| LMArena Text: medicine and healthcare | 1428 | 1342 | Gemini 2.5 Flash |
| LMArena Text: legal and government | 1430 | 1327 | Gemini 2.5 Flash |
| LMArena Text: business and finance | 1404 | 1311 | Gemini 2.5 Flash |
| Neither model has a result on the other 11 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | Gemini 2.5 Flash | Gemma 3 4B | Leader |
|---|---|---|---|
| Design Arena: websites | 1074 | Not measured | Only Gemini 2.5 Flash measured |
| Design Arena: UI components | 1005 | Not measured | Only Gemini 2.5 Flash measured |
| Design Arena: game dev | 1005 | Not measured | Only Gemini 2.5 Flash measured |
| Design Arena: data viz | 1060 | Not measured | Only Gemini 2.5 Flash measured |
| Design Arena: 3D | 1021 | Not measured | Only Gemini 2.5 Flash measured |
| Design Arena: all code categories | 1059 | Not measured | Only Gemini 2.5 Flash measured |
| Design Arena: SVG | 1020 | Not measured | Only Gemini 2.5 Flash measured |
| Neither model has a result on the other 2 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index150
| Benchmark | Gemini 2.5 Flash | Gemma 3 4B | Leader |
|---|---|---|---|
| SAGE (Vals AI) | 44.8% | Not measured | Only Gemini 2.5 Flash measured |
| MedScribe (Vals AI) | 83.0% | Not measured | Only Gemini 2.5 Flash measured |
| IFBench | 50.3% | 28.3% | Gemini 2.5 Flash |
| Long context reasoning (AA-LCR) | 65.3% | 6.7% | Gemini 2.5 Flash |
| DTBench | 76.5% | 50.9% | Gemini 2.5 Flash |
| LMCA | 27.5% | 2.8% | Gemini 2.5 Flash |
| LMArena Text | 1417 | 1291 | Gemini 2.5 Flash |
| LMArena Text: instruction following | 1405 | 1239 | Gemini 2.5 Flash |
| LMArena Text: creative writing | 1400 | 1270 | Gemini 2.5 Flash |
| LMArena Vision | 1235 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Text: English | 1418 | 1300 | Gemini 2.5 Flash |
| LMArena Text: non-English | 1409 | 1273 | Gemini 2.5 Flash |
| LMArena Text: Chinese | 1449 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Text: German | 1417 | 1281 | Gemini 2.5 Flash |
| LMArena Text: Japanese | 1396 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Text: Korean | 1385 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Text: French | 1432 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Text: Spanish | 1422 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Text: Russian | 1415 | 1294 | Gemini 2.5 Flash |
| LMArena Text: Polish | 1413 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Text: multi-turn | 1408 | 1255 | Gemini 2.5 Flash |
| LMArena Text: longer queries | 1419 | 1273 | Gemini 2.5 Flash |
| LMArena Text: writing and language | 1405 | 1288 | Gemini 2.5 Flash |
| LMArena Text: entertainment and media | 1393 | 1233 | Gemini 2.5 Flash |
| LMArena Vision: English | 1239 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Vision: Chinese | 1284 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Vision: reading text | 1237 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Vision: humor | 1225 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Vision: captioning | 1277 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Vision: recognizing things | 1252 | Not measured | Only Gemini 2.5 Flash measured |
| LMArena Vision: creative writing | 1241 | Not measured | Only Gemini 2.5 Flash measured |
| Neither model has a result on the other 15 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Gemini 2.5 Flash | Gemma 3 4B | 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
Gemma 3 4B costs less
Repository review
50K input + 3K output tokens
Gemma 3 4B costs less
Agent loop
200K input + 10K output tokens
Gemma 3 4B 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 | Google Gemini 2.5 Flash | Google Gemma 3 4B |
|---|---|---|
| Maker | ||
| Weights | No public weights listed | Open weights |
| Released | June 17, 2025 | March 13, 2025 |
| Context window | 1.05M | 131K |
| Input $/Mtok | $0.30 | $0.045 |
| Output $/Mtok | $2.50 | $0.075 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | $0.466 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Gemini 2.5 Flash better than Gemma 3 4B for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick Gemini 2.5 Flash for a larger context window (1.05M against 131K tokens).
Which is cheaper, Gemini 2.5 Flash or Gemma 3 4B?
Gemini 2.5 Flash costs $0.30 per million input tokens and $2.50 per million output tokens; Gemma 3 4B costs $0.045 and $0.075. Cost per solved task is known for Gemini 2.5 Flash only.
Which has the bigger context window, Gemini 2.5 Flash or Gemma 3 4B?
Gemini 2.5 Flash, with 1.05M tokens against 131K.
How many benchmarks is the Gemini 2.5 Flash vs Gemma 3 4B comparison based on?
36 benchmarks have a result for both models as of October 21, 2026. Gemini 2.5 Flash's coding index rests on 0 independent benchmarks and Gemma 3 4B's on 0. Where only one model has a result, the page says so instead of naming a winner.