Gemini 3.5 Flash Lite vs Gemma 3 4B
Gemini 3.5 Flash Lite leads Gemma 3 4B on the Artificial Analysis Intelligence Index, 22.2 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; 33 benchmarks have a result for both.
Gemini 3.5 Flash Lite or Gemma 3 4B: which to pick
Only Gemini 3.5 Flash Lite has a coding score, 49.0 on the Vibe Coding Index; Gemma 3 4B has none, so the two cannot be ranked on coding.
Pick Gemma 3 4B for output tokens 33.3x cheaper ($0.075 vs $2.50 per million).
- Shared results
- 33
- Gemini 3.5 Flash Lite only
- 52
- Gemma 3 4B only
- 9
- Like-for-like categories
- 6 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
Only Gemini 3.5 Flash Lite has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
Gemini 3.5 Flash Lite
Gemini 3.5 Flash Lite leads, 22.2 against 4.8.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only Gemini 3.5 Flash Lite has a AA Agentic Index result, 14.3.
Directional only
Reasoning
GPQA Diamond
No call
Only Gemini 3.5 Flash Lite has a GPQA Diamond result, 83.8%.
Directional only
Knowledge
MMLU Pro
No call
Only Gemini 3.5 Flash Lite has a MMLU Pro result, 85.8%.
Directional only
Long documents
Listed context window
Gemini 3.5 Flash Lite
Gemini 3.5 Flash Lite takes 1.05M tokens against 131K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Gemma 3 4B
Gemini 3.5 Flash Lite $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 3.5 Flash Lite has the number, $0.41.
Directional only
At a glance
- Coding index
- Gemini 3.5 Flash Lite 49.0 from 3 benchmarks, Gemma 3 4B not benchmarked.
- Widest category
- Math, 41.9 points apart on average over 3 shared benchmarks.
- Widest single gap
- Long context reasoning (AA-LCR): Gemini 3.5 Flash Lite 76.0% against Gemma 3 4B 6.7%, 69.3 points apart on a 0 to 100 scale.
- Output price
- Gemini 3.5 Flash Lite $2.50 against Gemma 3 4B $0.075 per million output tokens, 33.3x apart.
- Context window
- Gemini 3.5 Flash Lite 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 3.5 Flash Lite | Bars | Gemma 3 4B | Ties |
|---|---|---|---|---|
| Coding | 4 | 0 | ||
| Intelligence | 2 | 0 | ||
| Reasoning | 6 | 0 | ||
| Math | 3 | 0 | ||
| Knowledge | 4 | 0 | ||
| General | 14 | 0 | ||
| All shared boards | 33 | 0 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index40
| Benchmark | Gemini 3.5 Flash Lite | Gemma 3 4B | Leader |
|---|---|---|---|
| LMArena WebDev | 1435 | Not measured | Only Gemini 3.5 Flash Lite measured |
| AA Coding Index | 49.3 | 2.7 | Gemini 3.5 Flash Lite |
| LiveCodeBench (Vals AI) | 79.0% | Not measured | Only Gemini 3.5 Flash Lite measured |
| SWE-bench Verified (Vals AI) | 75.0% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Vibe Code Bench (Vals AI) | 37.2% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Code migration (Vals AI) | 6.2% | Not measured | Only Gemini 3.5 Flash Lite measured |
| ProgramBench (Vals AI) | 0.0% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Terminal-Bench 2.1 (Vals AI) | 50.2% | Not measured | Only Gemini 3.5 Flash Lite measured |
| CyberBench (Vals AI) | 59.1% | Not measured | Only Gemini 3.5 Flash Lite measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 11.2% | Only Gemma 3 4B measured |
| SciCode | 41.3% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 0.8% | Only Gemma 3 4B measured |
| WeirdML | 39.0% | Not measured | Only Gemini 3.5 Flash Lite measured |
| ALE-Bench | 765.3 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: coding | 1453 | 1230 | Gemini 3.5 Flash Lite |
| LMArena Text: software and IT | 1451 | 1269 | Gemini 3.5 Flash Lite |
| LMArena WebDev: React | 1408 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena WebDev: from an image | 1448 | Not measured | Only Gemini 3.5 Flash Lite measured |
| SciCode | 6.2% | 0.0% | Gemini 3.5 Flash Lite |
| Neither model has a result on the other 20 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | Gemini 3.5 Flash Lite | Gemma 3 4B | Leader |
|---|---|---|---|
| AA Intelligence Index | 22.2 | 4.8 | Gemini 3.5 Flash Lite |
| Epoch Capabilities Index | 144.9 | 115.9 | Gemini 3.5 Flash Lite |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | Gemini 3.5 Flash Lite | Gemma 3 4B | Leader |
|---|---|---|---|
| AA Agentic Index | 14.3 | Not measured | Only Gemini 3.5 Flash Lite measured |
| Finance Agent (Vals AI) | 47.4% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Tax Agent Bench (Vals AI) | 7.2% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Legal Research Bench (Vals AI) | 13.9% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Excel Modeling Benchmark (Vals AI) | 43.2% | Not measured | Only Gemini 3.5 Flash Lite measured |
| τ²-Bench Telecom | Not measured | 5.0% | Only Gemma 3 4B measured |
| APEX-Agents | 29.3% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Berkeley Function Calling Leaderboard (BFCL) | Not measured | 19.6% | Only Gemma 3 4B measured |
| Neither model has a result on the other 33 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | Gemini 3.5 Flash Lite | Gemma 3 4B | Leader |
|---|---|---|---|
| GPQA Diamond | 83.8% | Not measured | Only Gemini 3.5 Flash Lite measured |
| MMMU Pro (Vals AI) | 83.6% | Not measured | Only Gemini 3.5 Flash Lite measured |
| GPQA Diamond (Artificial Analysis run) | 83.8% | 29.1% | Gemini 3.5 Flash Lite |
| Humanity's Last Exam (Artificial Analysis run) | 18.8% | 5.3% | Gemini 3.5 Flash Lite |
| GPQA Diamond (Epoch AI run) | 83.3% | 23.2% | Gemini 3.5 Flash Lite |
| ARC-AGI | 53.5% | Not measured | Only Gemini 3.5 Flash Lite measured |
| ARC-AGI-2 | 10.3% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Chess puzzles | 21.0% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Mystery game puzzles | 19.0% | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: hard prompts | 1441 | 1253 | Gemini 3.5 Flash Lite |
| LMArena Text: expert | 1434 | 1223 | Gemini 3.5 Flash Lite |
| LMArena Text: hard prompts in English | 1440 | 1253 | Gemini 3.5 Flash Lite |
| LMArena Vision: diagrams | 1284 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Vision: homework | 1291 | Not measured | Only Gemini 3.5 Flash Lite measured |
| Kagi LLM Benchmark | Not measured | 25.2% | Only Gemma 3 4B measured |
| Neither model has a result on the other 7 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index30
| Benchmark | Gemini 3.5 Flash Lite | Gemma 3 4B | Leader |
|---|---|---|---|
| AA Math Index | Not measured | 12.7 | Only Gemma 3 4B measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 12.7% | Only Gemma 3 4B measured |
| OTIS Mock AIME | 71.1% | 7.5% | Gemini 3.5 Flash Lite |
| FrontierMath, tiers 1 to 3 | 26.0% | Not measured | Only Gemini 3.5 Flash Lite measured |
| FrontierMath, tier 4 | 0.0% | Not measured | Only Gemini 3.5 Flash Lite measured |
| ProofBench | 13.0% | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: math | 1434 | 1239 | Gemini 3.5 Flash Lite |
| LMArena Text: mathematical work | 1433 | 1256 | Gemini 3.5 Flash Lite |
| Neither model has a result on the other 5 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index40
| Benchmark | Gemini 3.5 Flash Lite | Gemma 3 4B | Leader |
|---|---|---|---|
| MMLU Pro | 85.8% | Not measured | Only Gemini 3.5 Flash Lite measured |
| LegalBench (Vals AI) | 84.1% | Not measured | Only Gemini 3.5 Flash Lite measured |
| CorpFin (Vals AI) | 60.7% | Not measured | Only Gemini 3.5 Flash Lite measured |
| TaxEval (Vals AI) | 72.6% | Not measured | Only Gemini 3.5 Flash Lite measured |
| MortgageTax (Vals AI) | 68.7% | Not measured | Only Gemini 3.5 Flash Lite measured |
| MedCode (Vals AI) | 43.5% | Not measured | Only Gemini 3.5 Flash Lite measured |
| MMLU Pro (Artificial Analysis run) | Not measured | 41.7% | Only Gemma 3 4B measured |
| LMArena Text: science | 1451 | 1306 | Gemini 3.5 Flash Lite |
| LMArena Text: medicine and healthcare | 1438 | 1342 | Gemini 3.5 Flash Lite |
| LMArena Text: legal and government | 1431 | 1327 | Gemini 3.5 Flash Lite |
| LMArena Text: business and finance | 1423 | 1311 | Gemini 3.5 Flash Lite |
| FACTS Benchmark Suite | 53.5% | Not measured | Only Gemini 3.5 Flash Lite measured |
| Neither model has a result on the other 6 boards here | |||
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Gemini 3.5 Flash Lite | Gemma 3 4B | Leader |
|---|---|---|---|
| Neither model has a result on 9 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index140
| Benchmark | Gemini 3.5 Flash Lite | Gemma 3 4B | Leader |
|---|---|---|---|
| Public Benefits Bench (Vals AI) | 51.8% | Not measured | Only Gemini 3.5 Flash Lite measured |
| SAGE (Vals AI) | 47.3% | Not measured | Only Gemini 3.5 Flash Lite measured |
| MedScribe (Vals AI) | 70.9% | Not measured | Only Gemini 3.5 Flash Lite measured |
| IFBench | Not measured | 28.3% | Only Gemma 3 4B measured |
| Long context reasoning (AA-LCR) | 76.0% | 6.7% | Gemini 3.5 Flash Lite |
| DTBench | 83.5% | 50.9% | Gemini 3.5 Flash Lite |
| LMCA | 37.6% | 2.8% | Gemini 3.5 Flash Lite |
| LMArena Text | 1435 | 1291 | Gemini 3.5 Flash Lite |
| LMArena Text: instruction following | 1419 | 1239 | Gemini 3.5 Flash Lite |
| LMArena Text: creative writing | 1421 | 1270 | Gemini 3.5 Flash Lite |
| LMArena Vision | 1269 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: English | 1438 | 1300 | Gemini 3.5 Flash Lite |
| LMArena Text: non-English | 1427 | 1273 | Gemini 3.5 Flash Lite |
| LMArena Text: Chinese | 1468 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: German | 1454 | 1281 | Gemini 3.5 Flash Lite |
| LMArena Text: Japanese | 1426 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: Korean | 1407 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: French | 1451 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: Spanish | 1441 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: Russian | 1443 | 1294 | Gemini 3.5 Flash Lite |
| LMArena Text: Polish | 1437 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Text: multi-turn | 1445 | 1255 | Gemini 3.5 Flash Lite |
| LMArena Text: longer queries | 1435 | 1273 | Gemini 3.5 Flash Lite |
| LMArena Text: writing and language | 1428 | 1288 | Gemini 3.5 Flash Lite |
| LMArena Text: entertainment and media | 1414 | 1233 | Gemini 3.5 Flash Lite |
| LMArena Vision: English | 1269 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Vision: Chinese | 1311 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Vision: reading text | 1280 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Vision: humor | 1287 | Not measured | Only Gemini 3.5 Flash Lite measured |
| LMArena Vision: creative writing | 1260 | Not measured | Only Gemini 3.5 Flash Lite 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 | Gemini 3.5 Flash Lite | 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 3.5 Flash Lite | Google Gemma 3 4B |
|---|---|---|
| Maker | ||
| Weights | No public weights listed | Open weights |
| Released | July 21, 2026 | March 13, 2025 |
| Context window | 1.05M | 131K |
| Input $/Mtok | $0.30 | $0.045 |
| Output $/Mtok | $2.50 | $0.075 |
| Coding index | 49.0 | Not benchmarked |
| Independent measurements behind it | 3 | 0 |
| Cost per solved task | $0.41 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Gemini 3.5 Flash Lite better than Gemma 3 4B for coding?
Only Gemini 3.5 Flash Lite has a coding score, 49.0 on the Vibe Coding Index; Gemma 3 4B has none, so the two cannot be ranked on coding. Pick Gemma 3 4B for output tokens 33.3x cheaper ($0.075 vs $2.50 per million).
Which is cheaper, Gemini 3.5 Flash Lite or Gemma 3 4B?
Gemini 3.5 Flash Lite 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 3.5 Flash Lite only.
Which has the bigger context window, Gemini 3.5 Flash Lite or Gemma 3 4B?
Gemini 3.5 Flash Lite, with 1.05M tokens against 131K.
How many benchmarks is the Gemini 3.5 Flash Lite vs Gemma 3 4B comparison based on?
33 benchmarks have a result for both models as of October 21, 2026. Gemini 3.5 Flash Lite's coding index rests on 3 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.