Enter opens the highlighted result. Esc closes.

Gemini 2.5 Pro vs Gemma 3 4B

Gemini 2.5 Pro leads Gemma 3 4B on the Artificial Analysis Intelligence Index, 16.1 to 4.8, and Gemma 3 4B costs 133.3x less per output token ($0.075 vs $10.00 per million). Updated October 21, 2026; 39 benchmarks have a result for both.

Replace Gemini 2.5 Pro Replace Gemma 3 4B
Gemini 2.5 ProGoogle
Coding
19.1
Overall
16.1
Output $/M
$10.00
39
Gemma 3 4BGoogle
Coding
No data
Overall
4.8
Output $/M
$0.075
0
vs boards won of 39 shared

Gemini 2.5 Pro or Gemma 3 4B: which to pick

Only Gemini 2.5 Pro has a coding score, 19.1 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 133.3x cheaper ($0.075 vs $10.00 per million).

Model AGemini 2.5 ProGoogle19.1/1003 independent benchmarks$0.64 per solved taskModel BGemma 3 4BGoogleNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
39
Gemini 2.5 Pro only
48
Gemma 3 4B only
0
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.

At a glance

Coding index
Gemini 2.5 Pro 19.1 from 3 benchmarks, Gemma 3 4B not benchmarked.
Widest category
Math, 59.1 points apart on average over 5 shared benchmarks.
Widest single gap
OTIS Mock AIME: Gemini 2.5 Pro 84.7% against Gemma 3 4B 7.5%, 77.2 points apart on a 0 to 100 scale.
Output price
Gemini 2.5 Pro $10.00 against Gemma 3 4B $0.075 per million output tokens, 133.3x apart.
Context window
Gemini 2.5 Pro 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.

Boards won, where both have a result
CategoryGemini 2.5 ProBarsGemma 3 4BTies
Coding50
Intelligence20
Agentic work10
Reasoning60
Math50
Knowledge50
General150
All shared boards390
CodingLike-for-likeSix of these boards feed the Vibe Coding Index50
Coding benchmarks for Gemini 2.5 Pro and Gemma 3 4B
BenchmarkGemini 2.5 ProGemma 3 4BLeader
SWE-bench Verified53.6%Not measuredOnly Gemini 2.5 Pro measured
LMArena WebDev1226Not measuredOnly Gemini 2.5 Pro measured
AA Coding Index33.32.7Gemini 2.5 Pro
SWE-bench Verified (Vals AI)54.4%Not measuredOnly Gemini 2.5 Pro measured
Vibe Code Bench (Vals AI)0.4%Not measuredOnly Gemini 2.5 Pro measured
LiveCodeBench (Artificial Analysis run)80.1%11.2%Gemini 2.5 Pro
SciCode46.3%Not measuredOnly Gemini 2.5 Pro measured
Terminal-Bench Hard (Artificial Analysis run)26.5%0.8%Gemini 2.5 Pro
Terminal-Bench32.6%Not measuredOnly Gemini 2.5 Pro measured
GSO-Bench3.9%Not measuredOnly Gemini 2.5 Pro measured
WeirdML54.0%Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: coding14521230Gemini 2.5 Pro
LMArena Text: software and IT14591269Gemini 2.5 Pro
LMArena WebDev: HTML1235Not measuredOnly Gemini 2.5 Pro measured
LMArena WebDev: from an image1285Not measuredOnly Gemini 2.5 Pro measured
Neither model has a result on the other 23 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
Intelligence benchmarks for Gemini 2.5 Pro and Gemma 3 4B
BenchmarkGemini 2.5 ProGemma 3 4BLeader
AA Intelligence Index16.14.8Gemini 2.5 Pro
Epoch Capabilities Index145.1115.9Gemini 2.5 Pro
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
Agentic work benchmarks for Gemini 2.5 Pro and Gemma 3 4B
BenchmarkGemini 2.5 ProGemma 3 4BLeader
AA Agentic Index1.6Not measuredOnly Gemini 2.5 Pro measured
τ²-Bench Telecom54.1%5.0%Gemini 2.5 Pro
GDPval23.3%Not measuredOnly Gemini 2.5 Pro measured
DeepResearch Bench41.5%Not measuredOnly Gemini 2.5 Pro measured
Vending-Bench 2$574Not measuredOnly Gemini 2.5 Pro measured
Neither model has a result on the other 34 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index60
Reasoning benchmarks for Gemini 2.5 Pro and Gemma 3 4B
BenchmarkGemini 2.5 ProGemma 3 4BLeader
GPQA Diamond (Artificial Analysis run)84.4%29.1%Gemini 2.5 Pro
Humanity's Last Exam (Artificial Analysis run)22.5%5.3%Gemini 2.5 Pro
GPQA Diamond (Epoch AI run)85.3%23.2%Gemini 2.5 Pro
ARC-AGI41.0%Not measuredOnly Gemini 2.5 Pro measured
ARC-AGI-24.9%Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: hard prompts14551253Gemini 2.5 Pro
LMArena Text: expert14521223Gemini 2.5 Pro
LMArena Text: hard prompts in English14531253Gemini 2.5 Pro
LMArena Vision: diagrams1274Not measuredOnly Gemini 2.5 Pro measured
LMArena Vision: homework1281Not measuredOnly Gemini 2.5 Pro measured
Neither model has a result on the other 11 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index50
Math benchmarks for Gemini 2.5 Pro and Gemma 3 4B
BenchmarkGemini 2.5 ProGemma 3 4BLeader
AA Math Index87.712.7Gemini 2.5 Pro
AIME 2025 (Artificial Analysis run)87.7%12.7%Gemini 2.5 Pro
OTIS Mock AIME84.7%7.5%Gemini 2.5 Pro
FrontierMath, tiers 1 to 324.6%Not measuredOnly Gemini 2.5 Pro measured
FrontierMath, tier 40.0%Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: math14501239Gemini 2.5 Pro
LMArena Text: mathematical work14541256Gemini 2.5 Pro
Neither model has a result on the other 6 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index50
Knowledge benchmarks for Gemini 2.5 Pro and Gemma 3 4B
BenchmarkGemini 2.5 ProGemma 3 4BLeader
CaseLaw (Vals AI)63.9%Not measuredOnly Gemini 2.5 Pro measured
CorpFin (Vals AI)60.8%Not measuredOnly Gemini 2.5 Pro measured
MortgageTax (Vals AI)68.9%Not measuredOnly Gemini 2.5 Pro measured
MedCode (Vals AI)50.6%Not measuredOnly Gemini 2.5 Pro measured
MMLU Pro (Artificial Analysis run)86.2%41.7%Gemini 2.5 Pro
LMArena Text: science14801306Gemini 2.5 Pro
LMArena Text: medicine and healthcare14761342Gemini 2.5 Pro
LMArena Text: legal and government14801327Gemini 2.5 Pro
LMArena Text: business and finance14391311Gemini 2.5 Pro
Neither model has a result on the other 8 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
Web design benchmarks for Gemini 2.5 Pro and Gemma 3 4B
BenchmarkGemini 2.5 ProGemma 3 4BLeader
Design Arena: websites1167Not measuredOnly Gemini 2.5 Pro measured
Design Arena: UI components1140Not measuredOnly Gemini 2.5 Pro measured
Design Arena: game dev1111Not measuredOnly Gemini 2.5 Pro measured
Design Arena: data viz1221Not measuredOnly Gemini 2.5 Pro measured
Design Arena: 3D1081Not measuredOnly Gemini 2.5 Pro measured
Design Arena: all code categories1152Not measuredOnly 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 index150
General benchmarks for Gemini 2.5 Pro and Gemma 3 4B
BenchmarkGemini 2.5 ProGemma 3 4BLeader
SAGE (Vals AI)41.9%Not measuredOnly Gemini 2.5 Pro measured
MedScribe (Vals AI)73.5%Not measuredOnly Gemini 2.5 Pro measured
IFBench48.7%28.3%Gemini 2.5 Pro
Long context reasoning (AA-LCR)69.0%6.7%Gemini 2.5 Pro
DTBench82.4%50.9%Gemini 2.5 Pro
LMCA34.8%2.8%Gemini 2.5 Pro
Creative writing (Lech Mazur)83.8%Not measuredOnly Gemini 2.5 Pro measured
LMArena Text14581291Gemini 2.5 Pro
LMArena Text: instruction following14371239Gemini 2.5 Pro
LMArena Text: creative writing14541270Gemini 2.5 Pro
LMArena Vision1263Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: English14581300Gemini 2.5 Pro
LMArena Text: non-English14521273Gemini 2.5 Pro
LMArena Text: Chinese1507Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: German14871281Gemini 2.5 Pro
LMArena Text: Japanese1461Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: Korean1434Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: French1473Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: Spanish1474Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: Russian14611294Gemini 2.5 Pro
LMArena Text: Polish1479Not measuredOnly Gemini 2.5 Pro measured
LMArena Text: multi-turn14531255Gemini 2.5 Pro
LMArena Text: longer queries14491273Gemini 2.5 Pro
LMArena Text: writing and language14541288Gemini 2.5 Pro
LMArena Text: entertainment and media14411233Gemini 2.5 Pro
LMArena Vision: English1261Not measuredOnly Gemini 2.5 Pro measured
LMArena Vision: Chinese1300Not measuredOnly Gemini 2.5 Pro measured
LMArena Vision: reading text1270Not measuredOnly Gemini 2.5 Pro measured
LMArena Vision: humor1262Not measuredOnly Gemini 2.5 Pro measured
LMArena Vision: captioning1304Not measuredOnly Gemini 2.5 Pro measured
LMArena Vision: recognizing things1283Not measuredOnly Gemini 2.5 Pro measured
LMArena Vision: creative writing1265Not measuredOnly Gemini 2.5 Pro measured
LMArena Document1421Not measuredOnly Gemini 2.5 Pro measured
Neither model has a result on the other 13 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Gemini 2.5 Pro and Gemma 3 4B
BenchmarkGemini 2.5 ProGemma 3 4BLeader
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

Gemini 2.5 Pro$0.0063Fits in one request
Gemma 3 4B$0.0001Fits in one request

Gemma 3 4B costs less

Repository review

50K input + 3K output tokens

Gemini 2.5 Pro$0.092Fits in one request
Gemma 3 4B$0.0025Fits in one request

Gemma 3 4B costs less

Agent loop

200K input + 10K output tokens

Gemini 2.5 Pro$0.350Fits in one request
Gemma 3 4B$0.0097Exceeds the context window

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.

Specifications of Gemini 2.5 Pro and Gemma 3 4B
SpecificationGoogle Gemini 2.5 ProGoogle Gemma 3 4B
MakerGoogleGoogle
WeightsNo public weights listedOpen weights
ReleasedJune 17, 2025March 13, 2025
Context window1.05M131K
Input $/Mtok$1.25$0.045
Output $/Mtok$10.00$0.075
Coding index19.1Not benchmarked
Independent measurements behind it30
Cost per solved task$0.64Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is Gemini 2.5 Pro better than Gemma 3 4B for coding?

Only Gemini 2.5 Pro has a coding score, 19.1 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 133.3x cheaper ($0.075 vs $10.00 per million).

Which is cheaper, Gemini 2.5 Pro or Gemma 3 4B?

Gemini 2.5 Pro costs $1.25 per million input tokens and $10.00 per million output tokens; Gemma 3 4B costs $0.045 and $0.075. Cost per solved task is known for Gemini 2.5 Pro only.

Which has the bigger context window, Gemini 2.5 Pro or Gemma 3 4B?

Gemini 2.5 Pro, with 1.05M tokens against 131K.

How many benchmarks is the Gemini 2.5 Pro vs Gemma 3 4B comparison based on?

39 benchmarks have a result for both models as of October 21, 2026. Gemini 2.5 Pro'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.

Compare next