Skip to content
VECTOR WIREAI INTELLIGENCE
UTC

Gemini 3.5 Flash vs Seed2.1

19 SHARED BENCHMARKS

Across 19 shared benchmarks, Gemini 3.5 Flash scores higher on 10 and Seed2.1 on 9. The widest gap is GDPVal, where Seed2.1 scores 82.7 against 34.2.

GOOGLEVSX19 SHARED10–9 HEAD-TO-HEADNEWEST SCORE ADDED

At a glance

MakerGooglex
Released—
Price per 1M tokens input / output$1.50 / $9.00—
Cost of 1M in + 1M out$10.50—
Head-to-head of 19 shared benchmarks10 wins9 wins
Scores tracked independently verified132 29 ◆50 0 ◆

Gemini 3.5 Flash's release date per Artificial Analysis. Prices: Artificial Analysis for Gemini 3.5 Flash. ◆ = independently verified score.

Where each leadsby capability area · benchmark wins

AreaGemini 3.5 FlashWINSSeed2.1
Reasoning10Gemini 3.5 Flash leads 1 of 1 · widest: ARC-AGI-2 72.1 vs 61.3
Coding11Even, 1–1 of 2
Agentic21Gemini 3.5 Flash leads 2 of 3 · widest: AA ApexAgents 47.1 vs 29.2
Multimodal11Even, 1–1 of 2

Biggest gaps

Gemini 3.5 Flash pulls furthest ahead on

  1. AA ApexAgents47.1 vs 29.2
  2. ARC-AGI-272.1 vs 61.3
  3. Terminal-Bench 2.178.7 vs 67.6

Seed2.1 pulls furthest ahead on

  1. GDPVal82.7 vs 34.2
  2. SciCode57.8 vs 53.9

Every shared benchmark19 · grouped by area

Reasoning 1

ARC-AGI-272.161.3

Coding 2

Agentic 3

GDPVal34.282.7
MCP Atlas83.680.3

Multimodal 2

MMMU-Pro84.380.1
Video-MME87.289

Other shared benchmarks 11

LVBench76.376.8
Minerva68.665.9
MotionBench70.674.8
OVBench56.569.7
TOMATO71.956.8
Toolathlon56.549.1
TVBench76.477.2
VideoHolmes67.167.6

Each score is the one the model's own page shows — the most authoritative tracked result for that benchmark, on the benchmark's standard methodology (source-attributed). ↓ marks lower-is-better metrics; ◆ an independently verified score. Open either model for its full surface, provenance and pricing.

Compare Gemini 3.5 Flash withALL PAIRINGS →

Compare Seed2.1 withALL PAIRINGS →