Enter opens the highlighted result. Esc closes.

DeepSeek V3.2 Exp vs Gemma 3 4B

DeepSeek V3.2 Exp leads Gemma 3 4B on the Artificial Analysis Intelligence Index, 16.6 to 4.8, and Gemma 3 4B costs 5.5x less per output token ($0.075 vs $0.41 per million). Updated October 21, 2026; 38 benchmarks have a result for both.

Replace DeepSeek V3.2 Exp Replace Gemma 3 4B
DeepSeek V3.2 ExpDeepSeek
Coding
No data
Overall
No data
Output $/M
$0.41
38
Gemma 3 4BGoogle
Coding
No data
Overall
22.6
Output $/M
$0.075
0
vs boards won of 38 shared

DeepSeek V3.2 Exp or Gemma 3 4B: which to pick

Neither model has a coding score yet, so the two cannot be ranked on coding.

Pick DeepSeek V3.2 Exp for a larger context window (164K against 131K tokens).

Model ADeepSeek V3.2 ExpDeepSeekNot benchmarked1 independent benchmarkNo cost per solved taskModel BGemma 3 4BGoogleNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
38
DeepSeek V3.2 Exp only
21
Gemma 3 4B only
4
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
DeepSeek V3.2 Exp not benchmarked, Gemma 3 4B not benchmarked.
Widest category
Math, 52.5 points apart on average over 4 shared benchmarks.
Widest single gap
AA Math Index: DeepSeek V3.2 Exp 87.7 against Gemma 3 4B 12.7, 75.0 points apart on a 0 to 100 scale.
Output price
DeepSeek V3.2 Exp $0.41 against Gemma 3 4B $0.075 per million output tokens, 5.5x apart.
Context window
DeepSeek V3.2 Exp takes 164K 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
CategoryDeepSeek V3.2 ExpBarsGemma 3 4BTies
Coding40
Intelligence20
Agentic work20
Reasoning60
Math40
Knowledge50
General150
All shared boards380
CodingLike-for-likeSix of these boards feed the Vibe Coding Index40
Coding benchmarks for DeepSeek V3.2 Exp and Gemma 3 4B
BenchmarkDeepSeek V3.2 ExpGemma 3 4BLeader
Aider polyglot74.2%Not measuredOnly DeepSeek V3.2 Exp measured
LMArena WebDev1272Not measuredOnly DeepSeek V3.2 Exp measured
AA Coding IndexNot measured2.7Only Gemma 3 4B measured
LiveCodeBench (Artificial Analysis run)78.9%11.2%DeepSeek V3.2 Exp
Terminal-Bench Hard (Artificial Analysis run)31.1%0.8%DeepSeek V3.2 Exp
WeirdML39.5%Not measuredOnly DeepSeek V3.2 Exp measured
LMArena Text: coding14381230DeepSeek V3.2 Exp
LMArena Text: software and IT14381269DeepSeek V3.2 Exp
LMArena WebDev: HTML1281Not measuredOnly DeepSeek V3.2 Exp measured
SciCodeNot measured0.0%Only Gemma 3 4B measured
Neither model has a result on the other 29 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
Intelligence benchmarks for DeepSeek V3.2 Exp and Gemma 3 4B
BenchmarkDeepSeek V3.2 ExpGemma 3 4BLeader
AA Intelligence Index16.64.8DeepSeek V3.2 Exp
Epoch Capabilities Index144.9115.9DeepSeek V3.2 Exp
Agentic workLike-for-likePublic boards, shown as published, kept out of the index20
Agentic work benchmarks for DeepSeek V3.2 Exp and Gemma 3 4B
BenchmarkDeepSeek V3.2 ExpGemma 3 4BLeader
τ²-Bench Telecom33.9%5.0%DeepSeek V3.2 Exp
TheAgentCompany42.9%Not measuredOnly DeepSeek V3.2 Exp measured
Berkeley Function Calling Leaderboard (BFCL)56.7%19.6%DeepSeek V3.2 Exp
Neither model has a result on the other 38 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index60
Reasoning benchmarks for DeepSeek V3.2 Exp and Gemma 3 4B
BenchmarkDeepSeek V3.2 ExpGemma 3 4BLeader
GPQA Diamond (Artificial Analysis run)79.7%29.1%DeepSeek V3.2 Exp
Humanity's Last Exam (Artificial Analysis run)14.9%5.3%DeepSeek V3.2 Exp
GPQA Diamond (Epoch AI run)Not measured23.2%Only Gemma 3 4B measured
LMArena Text: hard prompts14291253DeepSeek V3.2 Exp
LMArena Text: expert14221223DeepSeek V3.2 Exp
LMArena Text: hard prompts in English14441253DeepSeek V3.2 Exp
Kagi LLM Benchmark52.2%25.2%DeepSeek V3.2 Exp
Neither model has a result on the other 15 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index40
Math benchmarks for DeepSeek V3.2 Exp and Gemma 3 4B
BenchmarkDeepSeek V3.2 ExpGemma 3 4BLeader
AA Math Index87.712.7DeepSeek V3.2 Exp
AIME 2025 (Artificial Analysis run)87.7%12.7%DeepSeek V3.2 Exp
OTIS Mock AIMENot measured7.5%Only Gemma 3 4B measured
LMArena Text: math14221239DeepSeek V3.2 Exp
LMArena Text: mathematical work14331256DeepSeek V3.2 Exp
Neither model has a result on the other 8 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index50
Knowledge benchmarks for DeepSeek V3.2 Exp and Gemma 3 4B
BenchmarkDeepSeek V3.2 ExpGemma 3 4BLeader
MMLU Pro (Artificial Analysis run)85.0%41.7%DeepSeek V3.2 Exp
LMArena Text: science14431306DeepSeek V3.2 Exp
LMArena Text: medicine and healthcare14681342DeepSeek V3.2 Exp
LMArena Text: legal and government14191327DeepSeek V3.2 Exp
LMArena Text: business and finance14121311DeepSeek V3.2 Exp
Neither model has a result on the other 13 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
Web design benchmarks for DeepSeek V3.2 Exp and Gemma 3 4B
BenchmarkDeepSeek V3.2 ExpGemma 3 4BLeader
Design Arena: websites1179Not measuredOnly DeepSeek V3.2 Exp measured
Design Arena: UI components1167Not measuredOnly DeepSeek V3.2 Exp measured
Design Arena: game dev1144Not measuredOnly DeepSeek V3.2 Exp measured
Design Arena: data viz1152Not measuredOnly DeepSeek V3.2 Exp measured
Design Arena: 3D1148Not measuredOnly DeepSeek V3.2 Exp measured
Design Arena: all code categories1171Not measuredOnly DeepSeek V3.2 Exp measured
Design Arena: SVG1032Not measuredOnly DeepSeek V3.2 Exp measured
Neither model has a result on the other 2 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index150
General benchmarks for DeepSeek V3.2 Exp and Gemma 3 4B
BenchmarkDeepSeek V3.2 ExpGemma 3 4BLeader
IFBench54.1%28.3%DeepSeek V3.2 Exp
Long context reasoning (AA-LCR)72.3%6.7%DeepSeek V3.2 Exp
DTBench87.7%50.9%DeepSeek V3.2 Exp
LMCA29.1%2.8%DeepSeek V3.2 Exp
Fiction.LiveBench83.3%Not measuredOnly DeepSeek V3.2 Exp measured
CL-bench13.2%Not measuredOnly DeepSeek V3.2 Exp measured
CL-bench Life9.5%Not measuredOnly DeepSeek V3.2 Exp measured
LMArena Text14251291DeepSeek V3.2 Exp
LMArena Text: instruction following14011239DeepSeek V3.2 Exp
LMArena Text: creative writing14041270DeepSeek V3.2 Exp
LMArena Text: English14411300DeepSeek V3.2 Exp
LMArena Text: non-English14041273DeepSeek V3.2 Exp
LMArena Text: Chinese1461Not measuredOnly DeepSeek V3.2 Exp measured
LMArena Text: German14401281DeepSeek V3.2 Exp
LMArena Text: Japanese1331Not measuredOnly DeepSeek V3.2 Exp measured
LMArena Text: Korean1372Not measuredOnly DeepSeek V3.2 Exp measured
LMArena Text: French1434Not measuredOnly DeepSeek V3.2 Exp measured
LMArena Text: Spanish1441Not measuredOnly DeepSeek V3.2 Exp measured
LMArena Text: Russian14121294DeepSeek V3.2 Exp
LMArena Text: Polish1435Not measuredOnly DeepSeek V3.2 Exp measured
LMArena Text: multi-turn14221255DeepSeek V3.2 Exp
LMArena Text: longer queries14181273DeepSeek V3.2 Exp
LMArena Text: writing and language14071288DeepSeek V3.2 Exp
LMArena Text: entertainment and media14071233DeepSeek V3.2 Exp
Neither model has a result on the other 22 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for DeepSeek V3.2 Exp and Gemma 3 4B
BenchmarkDeepSeek V3.2 ExpGemma 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

DeepSeek V3.2 Exp$0.0005Fits in one request
Gemma 3 4B$0.0001Fits in one request

Gemma 3 4B costs less

Repository review

50K input + 3K output tokens

DeepSeek V3.2 Exp$0.015Fits in one request
Gemma 3 4B$0.0025Fits in one request

Gemma 3 4B costs less

Agent loop

200K input + 10K output tokens

DeepSeek V3.2 Exp$0.058Exceeds the context window
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 DeepSeek V3.2 Exp and Gemma 3 4B
SpecificationDeepSeek V3.2 ExpGoogle Gemma 3 4B
MakerDeepSeekGoogle
WeightsOpen weightsOpen weights
ReleasedSeptember 29, 2025March 13, 2025
Context window164K131K
Input $/Mtok$0.27$0.045
Output $/Mtok$0.41$0.075
Coding indexNot benchmarkedNot benchmarked
Independent measurements behind it10
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is DeepSeek V3.2 Exp better than Gemma 3 4B for coding?

Neither model has a coding score yet, so the two cannot be ranked on coding. Pick DeepSeek V3.2 Exp for a larger context window (164K against 131K tokens).

Which is cheaper, DeepSeek V3.2 Exp or Gemma 3 4B?

DeepSeek V3.2 Exp costs $0.27 per million input tokens and $0.41 per million output tokens; Gemma 3 4B costs $0.045 and $0.075. Cost per solved task is known for neither of them.

Which has the bigger context window, DeepSeek V3.2 Exp or Gemma 3 4B?

DeepSeek V3.2 Exp, with 164K tokens against 131K.

How many benchmarks is the DeepSeek V3.2 Exp vs Gemma 3 4B comparison based on?

38 benchmarks have a result for both models as of October 21, 2026. DeepSeek V3.2 Exp's coding index rests on 1 independent benchmark 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