Noometry

Model comparison

DeepSeek-V3 vs Gemini 3.1 Pro Preview

Gemini 3.1 Pro Preview is the stronger model overall, scoring 56.7 to 39.5 on the Noometry Index. DeepSeek-V3 costs 11× less per token, which makes it the better buy when Gemini 3.1 Pro Preview's lead doesn't matter for your workload.

Last verified . 31 shared benchmarks.

DeepSeek-V3 DeepSeek

39.5

Rank #166 Confirmed

Gemini 3.1 Pro Preview Google

56.7

Rank #23 Confirmed

Summary

  • They share 31 benchmarks with published results for both. DeepSeek-V3 scores higher in 0 categories and Gemini 3.1 Pro Preview in 8 categories; 7 gaps are clear of the uncertainty.
  • The widest gap is in reasoning, where Gemini 3.1 Pro Preview leads 71.7 to 20.5.
  • The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 95.6% for Gemini 3.1 Pro Preview.
  • DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $2 / $12 for Gemini 3.1 Pro Preview.
  • Gemini 3.1 Pro Preview accepts more context: 1.05M tokens versus 164K.
  • DeepSeek-V3 has downloadable open weights; the other is API-only.

Side by side

DeepSeek-V3Gemini 3.1 Pro Preview
ProviderDeepSeekGoogle
Noometry Index39.556.7
Released2024-12-262026-02-19
WeightsOpenProprietary
Context window164K1.05M
Max output164K66K
Input $ / M tokens$0.24$2
Output $ / M tokens$0.90$12
Results tracked6071

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding Too close to call

DeepSeek-V3: 42.3 (#106), Gemini 3.1 Pro Preview: 42.5 (#99)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
SciCode35.8%58.9%
WeirdML36.1%72.1%
LMArena Coding13681484
SWE-bench Verified—75.6%
DeepSWE—11.7%
Aider Polyglot55.1%—
LMArena WebDev—1447
GSO—22.6%
BigCodeBench Instruct50%—
LiveBench Coding70.9%—
MirrorCode—8.9%
BigCodeBench Complete62.2%—
ALE-Bench—1,161
AlgoTune—2.02
HumanEval+86.6%—
MBPP+73%—

Agentic & Tool Use Not comparable

DeepSeek-V3: —, Gemini 3.1 Pro Preview: 37.7 (#34)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
METR Time Horizons49.6%77%
Terminal-Bench—80.2%
APEX-Agents—35.3%
τ²-bench Banking—26%
DeepResearch Bench—47.8%
PostTrainBench—22%
BALROG—57%
ExploitBench—26.1%
GBAEval—0.8%
GDP.pdf—17%
LMArena Search—1211
Vending-Bench 2—3,774

Reasoning Gemini 3.1 Pro Preview leads

DeepSeek-V3: 20.5 (#236), Gemini 3.1 Pro Preview: 71.7 (#12)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
SimpleBench27.2%79.6%
CritPt0%17.7%
LMArena Hard Prompts13651485
DTBench64.8%97.1%
LMCA15.5%53.8%
Epoch Capabilities Index135.94154.77
ForecastBench59.159
ARC-AGI-2—77.1%
Kagi LLM Benchmark52.3%—
NYT Connections (extended)—97.4%
ARC-AGI-1—98%
Chess Puzzles—55%
EnigmaEval—36.8%
Thematic Generalization—79.4%
EBR-Bench—14.3%
LiveBench Reasoning65.8%—
Mystery Game Puzzles—34%
LiveBench Data Analysis60.9%—
BIG-Bench Hard87.5%—
HellaSwag88.9%—
LiveBench66.9%—
PIQA84.7%—
WinoGrande85.2%—

Math Gemini 3.1 Pro Preview leads

DeepSeek-V3: 32.1 (#219), Gemini 3.1 Pro Preview: 62.1 (#34)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
OTIS Mock AIME 2024-202537.8%95.6%
LMArena Math13731485
FrontierMath (Feb 2025 set)1.7%36.9%
FrontierMath (Tiers 1-3)—59.6%
FrontierMath Tier 4—26.8%
MathArena Final-Answer Competitions—86.5%
ProofBench—26%
Omni-MATH40.3%—
LiveBench Math73.5%—
MATH Level 575.5%—
FrontierMath Tier 4 (v1)—16.7%

Knowledge Gemini 3.1 Pro Preview leads

DeepSeek-V3: 37.5 (#155), Gemini 3.1 Pro Preview: 71.8 (#3)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
GPQA Diamond67.6%94.4%
Vectara Hallucination Rate6.1%10.4%
LMArena Expert13511485
Humanity's Last Exam—46.4%
SimpleQA Verified—73.5%
MMLU-Pro72.3%—
Confabulations26.1%—
GPQA (HELM)53.8%—
ARC (AI2) Challenge95.3%—
MMLU87.2%—
TriviaQA82.9%—

Multimodal Not comparable

DeepSeek-V3: —, Gemini 3.1 Pro Preview: 37.9 (#69)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
LMArena Vision—1296
Blueprint-Bench 2—26.5%
Furniture Assembly—26.7%
LMArena Document—1444

Multilingual Gemini 3.1 Pro Preview leads

DeepSeek-V3: 48.5 (#143), Gemini 3.1 Pro Preview: 57.0 (#12)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
LMArena Non-English13581477
LMArena Chinese13911529
LMArena French13851487
LMArena German13741491
LMArena Japanese13331493
LMArena Korean13191455
LMArena Russian13731498
LMArena Spanish13581479

Instruction Following Gemini 3.1 Pro Preview leads

DeepSeek-V3: 72.8 (#130), Gemini 3.1 Pro Preview: 77.0 (#32)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
LMArena Instruction Following13451466
LiveBench Instruction Following81.5%—
IFEval83.2%—

Long Context Gemini 3.1 Pro Preview leads

DeepSeek-V3: 34.0 (#253), Gemini 3.1 Pro Preview: 47.4 (#18)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
LMArena Longer Query13521483
Fiction.LiveBench50%—
CL-bench—20.8%
CL-bench Life—16.9%

Writing & Preference Gemini 3.1 Pro Preview leads

DeepSeek-V3: 57.4 (#130), Gemini 3.1 Pro Preview: 66.1 (#37)

BenchmarkDeepSeek-V3Gemini 3.1 Pro Preview
LMArena Text13751481
LMArena Creative Writing13641482
EQ-Bench Creative Writing14721491
LMArena Multi-Turn13891488
Short-Story Creative Writing77%—
WildBench83%—
EQ-Bench 4—1142
LiveBench Language49.1%—

Frequently asked questions

Is DeepSeek-V3 better than Gemini 3.1 Pro Preview?

Gemini 3.1 Pro Preview is the stronger model overall, scoring 56.7 to 39.5 on the Noometry Index. DeepSeek-V3 costs 11× less per token, which makes it the better buy when Gemini 3.1 Pro Preview's lead doesn't matter for your workload.

Which is cheaper, DeepSeek-V3 or Gemini 3.1 Pro Preview?

DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Gemini 3.1 Pro Preview lists at $2 and $12.

Is DeepSeek-V3 or Gemini 3.1 Pro Preview better for coding?

They score almost the same on coding (42.3 vs 42.5); test both on your own repository before choosing.

Which has the bigger context window?

Gemini 3.1 Pro Preview does, with 1.05M tokens against 164K.

How many benchmarks do DeepSeek-V3 and Gemini 3.1 Pro Preview share?

31 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Gemini 3.1 Pro Preview has 71.

Related comparisons

Go deeper