Codesota · Models · Gemini 2.5 ProGoogle16 results · 15 benchmarks
Model card

Gemini 2.5 Pro.

GoogleapiMultimodal LLMProprietary3 current SOTA

#1 on OCRBench v2 Chinese, MME-VideoOCR

§ 02 · Benchmarks

Every benchmark Gemini 2.5 Pro has a recorded score for.

#BenchmarkArea · TaskMetricValueRankDateSource
01OCRBench v2Computer Vision · General OCR Capabilitiesoverall-zh-private62.2%#1/52025-03-25source ↗
02ARC-AGI-2Reasoning · Logical Reasoningaccuracy5.0%#1/3—source ↗
03MME-VideoOCRComputer Vision · General OCR Capabilitiestotal-accuracy73.7%#1/6—source ↗
04ARC-ChallengeReasoning · Commonsense Reasoningaccuracy97.8%#2/10—source ↗
05ThaiOCRBenchComputer Vision · Optical Character Recognitionted-score0.8%#2/5—source ↗
06AIME 2024Reasoning · Mathematical Reasoningaccuracy92.0%#3/11—source ↗
07OCRBench v2Computer Vision · General OCR Capabilitiesoverall-en-private59.3%#4/272025-03-25source ↗
08ARC-AGI-1Reasoning · Logical Reasoningaccuracy56.1%#4/5—source ↗
09GSM8KReasoning · Mathematical Reasoningaccuracy99.0%#4/48—source ↗
10MATHReasoning · Mathematical Reasoningaccuracy97.3%#6/46—source ↗
11AIME 2025Reasoning · Mathematical Reasoningaccuracy86.7%#12/22—source ↗
12OmniDocBenchComputer Vision · Document Parsingcomposite88.0%#14/34—source ↗
13MMLUReasoning · Commonsense Reasoningaccuracy89.8%#17/642025-06-17source ↗
14GPQA DiamondReasoning · Multi-step Reasoningaccuracy84.0%#25/74—source ↗
15SWE-Bench VerifiedComputer Code · Code Generationresolve-rate63.8%#25/39—source ↗
16SWE-bench VerifiedAgentic AI · SWE-benchresolve-rate63.2%#53/81—source ↗
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 03 · Strengths by area

Where Gemini 2.5 Pro actually performs.

Computer Vision
4
benchmarks
avg rank #4.4 · 2 SOTA
Reasoning
9
benchmarks
avg rank #8.2 · 1 SOTA
Computer Code
1
benchmark
avg rank #25.0
Agentic AI
1
benchmark
avg rank #53.0
§ 05 · Related models

Other Google models scored on Codesota.

Gemini 1.5 Pro
21 results
Gemini 3 Pro
Undisclosed params · 13 results
Gemini-3.1-Pro
11 results
Gemma-3-27b
27B params · 11 results
Gemma 3 (27B, IT)
9 results
Gemma-2-27b-it
9 results
gemma-3-12b-it
9 results
gemma-3-1b-it
9 results
§ 06 · Sources & freshness

Where these numbers come from.

google-technical-report
7
results
alphaxiv-leaderboard
4
results
arcprize-leaderboard
1
result
artificialanalysis
1
result
AlphaXiv
1
result
google-blog
1
result
editorial
1
result
9 of 16 rows marked verified. · first result 2025-03-25, latest 2025-06-17.