Gemini 3.1 Pro Preview

Overall ? Overall score composition: Reasoning 39% — 77.1; Code 20% — 93.3; Image 12% — 84.0; Text 24% — 100.0; Math 5% — 67.0

86.7

#16

BenchLM → Airating calculation

Code ? Airating rank — aggregate for Code (BenchLM / Airating calculation)

93.3

#7

BenchLM, Chatbot Arena

BenchLM · Code — 93.3 (confidence 81%); 1447 — 23.1 (confidence 95%)

Text ? Airating rank — aggregate for Text (BenchLM / Airating calculation)

100.0

#7

BenchLM, Chatbot Arena

BenchLM · Text — 100.0 (confidence 81%); 1486 — 63.3 (confidence 95%); 1486 — 63.3 (confidence 95%)

Image ? Airating rank — aggregate for Image (BenchLM / Airating calculation)

84.0

#45

BenchLM, Chatbot Arena

BenchLM · Image — 84.0 (confidence 81%); 1280 — 37.0 (confidence 95%)

Math ? Airating rank — aggregate for Math (BenchLM / Airating calculation)

67.0

#87

BenchLM

BenchLM · Math — 67.0 (confidence 81%)

Reasoning ? Based on benchmarks: ARC-AGI-2: 77.1%

77.1

#112

ARC Prize, BenchLM

ARC-AGI-2 — 77.1% (confidence 95%); BenchLM · Reasoning — 96.4 (confidence 81%)

Speed ? Output throughput (tokens/sec) from modelgrep live probes, or catalog estimate as fallback.

5.8

#304

Artificial Analysis · 8 hours ago

Price ? Normalized API cost (input + output per 1M tokens) from OpenRouter and models.dev.

81.5

#543

models.dev · 8 hours ago