Gemini 2.5 Pro

Gemini 2.5 Pro is a non-reasoning model from Google. 21 benchmarks count toward its score, in 6 categories.

availableShows if the model has enough results for an index.
IndexOverall score out of 100.46.2 ±5.0
CoverageShare of the index weight with results.85%
SpeedOutput tokens per second.87/s
Input / 1MUS dollars per 1M input tokens.$1.25 batch $0.625
Output / 1MUS dollars per 1M output tokens.$10 batch $5 US dollars per 1M output tokens in a batch.
ContextMaximum tokens in one request.1.05M
EloLMArena rating and rank.1458 (#36)

The index is a score out of 100. The ± range shows how much it can change. Batch work costs less.

122,554 votes. Elo shows what people prefer. It does not change the score.

CapabilitiesScore per category, out of 100.

Out of 100
AgenticMulti-step tasks with tools.
39.9
CodingCode writing and repair.
46.5
ReasoningLogic problems and puzzles.
49.9
MultimodalTasks with images and text.
58.6
KnowledgeFacts and expert knowledge.
51.4
MultilingualTasks in many languages.
N/A
InstructionTasks with strict rules in the prompt.
38.5
MathMath problems.
N/A

Results

21 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result as a score out of 100.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
Artificial Analysis GPQA DiamondKnowledge84.4%55.0Artificial Analysis
Graduate-Level Google-Proof Q&AKnowledge83.0%54.9David Rein et al.
Artificial Analysis MMMU-ProMultimodal74.9%58.6Artificial Analysis
Artificial Analysis Long Context ReasoningReasoning69.0%56.0Artificial Analysis
Software Engineering Benchmark VerifiedCoding63.8%48.7Carlos E. Jimenez et al.
SWE-benchCoding54.4%41.21 Sept 2026Vals AI
τ²-Bench Tool-Agent-User EvaluationAgentic54.1%37.6Victor Barres et al.
SWE-bench VerifiedCoding53.6%40.5mini-SWE-agent26 Feb 2026SWE-bench team
SWE-bench VerifiedCoding53.6%40.5mini-SWE-agent1 Sept 2026SWE-bench team
Artificial Analysis IFBenchInstruction48.7%38.5Artificial Analysis
Artificial Analysis SciCodeCoding46.3%57.1Artificial Analysis
Gert Labs Composite Game BenchmarkAgentic42.0%49.8Gert Labs
Terminal-Bench 1.0Agentic41.3%46.512 Jan 2026Vals AI
Artificial Analysis Omniscience AccuracyKnowledge39.1%62.2Artificial Analysis
Artificial Analysis Coding IndexCoding33.3%42.4Artificial Analysis
Terminal-Bench 2.0Agentic30.3%44.34 Jun 2026Vals AI
Artificial Analysis Humanity's Last ExamKnowledge22.5%49.0Artificial Analysis
Humanity's Last ExamKnowledge18.8%44.7Center for AI Safety et al.
IOI v1Coding17.1%49.69 Aug 2026Vals AI
Artificial Analysis Intelligence IndexKnowledge16.1%42.5Artificial Analysis
τ²-bench BankingAgentic13.7%8.6high effort · Sierra4 Aug 2026Sierra Research
Artificial Analysis Agentic IndexAgentic3.5%40.0Artificial Analysis
Critical Physics TasksReasoning2.6%43.8Artificial Analysis
Vibe Code Bench v1.1Coding0.4%42.3OpenHands21 Sept 2026Vals AI
GDPval-AA normalizedAgentic0.0%35.6Artificial Analysis

21 benchmarks count, from 25 of 25 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authorsOpenRouter, collected directlyNo licence statedVals AI, collected directlyNo licence stated. Read from the public leaderboard and credited to Vals AISWE-bench team, collected directlyNo licence stated for the leaderboard. The harness repo is MITSierra Research, collected directlyMIT — results are in the licensed repository

Same level, lower price

Qwen3.8-Flash-Next66.4 · Freedots3-note Preview65.5 · FreeApodex 1.1 Mini59.1 · Free

More from Google

Gemini 3.7 Flash72.3Gemini 3.8 Flash72.1Gemini 3.5 Flash67.3Gemini 3.6 Flash66.8Gemini 3.1 Pro64.7