gemini-2.5-pro

Language Proprietary

Benchmarks

About gemini-2.5-pro

gemini-2.5-pro is an AI model developed by Google DeepMind, in the language category, released in 2025, made available as a proprietary (API-only) model. Gemini 2.5 Pro release date (per deprecations table)..

On this page you'll find gemini-2.5-pro's full specifications, including 78 benchmark results. Review provider pricing and benchmark scores below, or compare gemini-2.5-pro head-to-head with other language models.

→ More models by Google DeepMind → All language models → AI benchmarks leaderboard

Links

Announcement ↗

Community

Would you run gemini-2.5-pro again?

No votes yet — be the first.

Did it run? — community reports by hardware

No reports yet. Own this model on some hardware? Be the first to confirm it runs.

Discussion

No comments yet. Start the discussion.

gemini-2.5-pro benchmark scores▶

Compare on the benchmark table →

Benchmark	Category	Score	Variant
BrowseComp	agentic	7.6%
TAU-bench Airline	agentic	60%
TAU-bench Retail	agentic	77%	reported in GLM-4.5 paper
Terminal-Bench 2.0	agentic	25.3%
Codeforces Elo	code	1,837	rating-cited-ring-1t
Codeforces Elo	code	1,675	lowthink-rating-cited-ling-1t
FullStackBench	code	48.19%	lowthink-cited-ling-1t
LiveCodeBench	code	80.1%	2407-2501
LiveCodeBench	code	45.43%	lowthink-cited-ling-1t
LiveCodeBench v5	code	69.1%	cited-phi4r
LiveCodeBench v5	code	58.24%
LiveCodeBench v6	code	70.65%	cited-ring-1t
MBPP	code	91.01%	lowthink-sanitized-cited-ling-1t
SWE-bench Verified	code	49%
SciCode	code	42.8%
Aider Benchmark	coding	94.36	cited-ring-1t
Aider Benchmark	coding	89.85	lowthink-edit-cited-ling-1t
MultiPL-E	coding	71.48	lowthink-cited-ling-1t
ArenaHard v2	general	79.2	win-rate-cited-ring-1t
ArenaHard v2	general	74.46	lowthink-win-rate-cited-ling-1t
C-Eval	general	90.14%	cited-ring-1t
C-Eval	general	88.77%	lowthink-cited-ling-1t
IFEval	general	90.8%
IFEval	general	88.72%	cited-ring-1t
IFEval	general	87.08%	lowthink-prompt-strict-cited-ling-1t
MMLU	general	91.9%
MMLU-Pro	general	86.2%
MMLU-Pro	general	85.62%	cited-ring-1t
MMLU-Pro	general	82.13%	lowthink-cited-ling-1t
MMLU-Redux	general	94.67%	lowthink-cited-ling-1t
MultiChallenge	general	57.5%
MultiChallenge	general	54.58%	cited-ring-1t
MultiChallenge	general	51.28%	lowthink-cited-ling-1t
AGIEval	knowledge	88.99	cited-ring-1t
CMMLU	knowledge	88.83	cited-ring-1t
TriviaQA	knowledge	84.45	cited-ring-1t
AIME 2024	math	92%	cited-phi4r
AIME 2024	math	88.7%
AIME 2024	math	87.24%
AIME 2024	math	77.5%	lowthink-cited-ling-1t
AIME 2025	math	88%	cited-ring-1t
AIME 2025	math	86.7%	cited-phi4r
AIME 2025	math	85.75%
AIME 2025	math	70.1%	lowthink-cited-ling-1t
CNMO 2024	math	80.64%	cited-ring-1t
CNMO 2024	math	74.65%	lowthink-cited-ling-1t
HMMT February 2025	math	82.5%	cited-phi4r
HMMT February 2025	math	82.5%	cited-ring-1t
HMMT February 2025	math	74.18%
HMMT February 2025	math	60.73%	lowthink-cited-ling-1t
MATH-500	math	96.7%
Omni-MATH	math	82.14	cited-ring-1t
Omni-MATH	math	72.02	lowthink-cited-ling-1t
Omni-MATH-HARD	math	69.36%
UGMathBench	math	74.5	cited-ring-1t
UGMathBench	math	70.1	lowthink-cited-ling-1t
ARC-AGI-1	reasoning	45.44	cited-ring-1t
ARC-AGI-1	reasoning	18.94	lowthink-cited-ling-1t
BBEH	reasoning	51.51	cited-ring-1t
BBEH	reasoning	29.08	lowthink-cited-ling-1t
FinanceReasoning	reasoning	87.33	cited-ring-1t
FinanceReasoning	reasoning	86.65	lowthink-cited-ling-1t
GPQA Diamond	reasoning	86.4%	cited-ring-1t
GPQA Diamond	reasoning	84.4%
GPQA Diamond	reasoning	84%	cited-phi4r
GPQA Diamond	reasoning	71.81%	lowthink-cited-ling-1t
Humanity's Last Exam	reasoning	21.6%	cited-ring-1t
Humanity's Last Exam	reasoning	21.1%
Humanity's Last Exam	reasoning	19.93%
Humanity's Last Exam	reasoning	12.07%	lowthink-cited-ling-1t
ZebraLogic	reasoning	92.4%	cited-ring-1t
ZebraLogic	reasoning	70.2%	lowthink-cited-ling-1t
HealthBench	safety	49.39%	cited-ring-1t
SimpleQA	safety	54%
PhyBench	science	55.01	cited-ring-1t
BFCL v3	tool-use	63.31%	lowthink-cited-ling-1t
BFCL v3	tool-use	61.36%	cited-ring-1t
BFCL v3	tool-use	61.2%

More models like gemini-2.5-pro

Llama-3-70B-Synthia-v3.5

· 70B

Frequently asked questions

What is gemini-2.5-pro? ▶

Gemini 2.5 Pro release date (per deprecations table). It is tracked on GenAIList with its specifications, benchmark scores and provider pricing.

Who created gemini-2.5-pro? ▶

gemini-2.5-pro was developed by Google DeepMind and released in 2025.

Is gemini-2.5-pro open source or proprietary? ▶

gemini-2.5-pro is a proprietary model. It is accessed through an API rather than by downloading the weights.

How much does gemini-2.5-pro cost? ▶

Pricing for gemini-2.5-pro depends on the provider. See the providers table on this page for the latest API rates.

How does gemini-2.5-pro perform on benchmarks? ▶

gemini-2.5-pro is benchmarked across 78 evaluations on GenAIList, including BrowseComp (7.6%). See the full benchmark table below and compare it with other models.