Gemini 2.5 Pro

+ contribute

Gemini 2.5 Pro is a closed-weights language model from Google, released June 2025.

67.2±0.6#110 of 719 LLMsfrom 52 benchmark results across 22 sources

$ how it compares

The LLMs ranked either side of Gemini 2.5 Pro, out of 828 with a score, then more from Google.

#	model	score	where	RAM	released
123	Hy3 preview (295B-A21B) Tencent	68±1	local	~629GB	2026-04
124	GPT-5 mini OpenAI	67±1	api $0.69/M		2025-08
125	Gemini 2.5 Pro Google	67±1	api $3.44/M		2025-06
126	Claude Opus 4 Anthropic	67±1	comparison only		2025-05
127	K-EXAONE 2.0 (750B-A37B) LG AI Research	67±1	local	~791GB	2026-07
	more from Google
9	Gemini 4 Argon (limited preview) Google	82±1	limited preview		2026-09
23	Gemini 3.8 Flash Google	77±1	api $1.50/M		2026-09
32	Gemini 3.7 Flash Google	76±1	api $1.50/M		2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #134 of 300 · 21% of its score · 12 tests

Trails SmolLM3 3B, Jamba 1.5 Large, GPT-4o (2024-05-13) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
SWE-bench Verified	57.6% ✎	▁··▁·▃▂▁▄█▄▁ beats 2/29	hard · avg 19%	▇▇▇ 6.8%	sharp · at its level
ContractBench success rate (33 observation-contract tool tasks)	51.5% ✎	▆··▆▄▂▄▂▆▆█▄ beats 13/23	very hard · avg 14%	▇ 2.7%	sharp · at its level
Terminal-Bench 2.0 with Terminus 2 agent (benchmark authors' runs)	32.6% ✎	▃··▃·▃▃██▅▃▅ beats 9/14	very hard · avg 5.8%	▇ 2.7%	sharp · at its level
τ²-bench Retail	70.9% ✎	█···█·█·██·█ beats 3/5	medium · avg 36%	▇ 2.2%	at its level
τ³-bench banking_knowledge (v1.0.1 grading)	13.7% ✎	▃▃▂▃·▆▃▂█▅▅▂ beats 3/25	very hard · avg 4.1%	▇ 1.5%	typical
τ²-bench Telecom	38.3% ✎	█··██·█·█··█ beats 1/5	very hard · avg 10%	▇ 1.4%	sharp · at its level · noisy
τ²-bench Airline	58.0% ✎	▄·····█··▄·█ beats 1/5	hard · avg 29%	▇ 1.1%	at its level
EEGAgentBench overall (tool-based EEG analysis agents, 6 tasks)	44.2% ✎	▂·▁▁··▇▇█▂·▁ beats 8/25	hard · avg 29%	▇ 0.8%	blunt · at its level

4 more tests

test	result	vs the others in its run	difficulty	counts	why
BFCL V4 (2025-11-03 archive, models absent from the December archive)	52.3% ✎	▃··▅▃▃·▅·▃·█ beats 6/10	hard · avg 34%	▇ 0.6%	at its level
Galileo Agent Leaderboard v2 (action completion)	43.0% ✎	▅▃·▃▃▅▅█▃█▅▅ beats 10/19	hard · avg 27%	▇ 0.6%	at its level
τ²-Bench-Verified (GPT-5.1 user)	56.4% ✎	▂··▄···▂▄▄█▄ beats 2/13	medium · avg 45%	▇ 0.6%	at its level
BFCL v3	63.2% ✎	▁▁▁▁▁▃▅█▆▅▄▁ beats 111/170	medium · avg 56%	▇ 0.3%	blunt · at its level

coding▁▁▁▃▆██▇▆▄▃▁ #164 of 458 · 16% of its score · 4 tests

Trails QwQ-32B-Preview, Ministral 3 8B Reasoning 2512, Phi-4-reasoning 14B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
SWE-bench Verified	57.6% ✎	▁··▁·▃▂▁▄█▄▁ beats 2/29	hard · avg 19%	▇▇▇ 6.8%	sharp · at its level
Arena WebDev (Code Arena)	1227.0 ✎	▁▂▄▃▅█▄▇▆▄▁▁ beats 7/119	avg 981	▇▇ 5.1%	sharp
Luau Leetcode fully solved (Torpedo Software)	65.4% ✎	▁▁▁▁·▁▂▃▅▃█▇ beats 18/49	hard · avg 18%	▇ 2.5%	at its level
LiveCodeBench (version unspecified)	74.2% ✎	▄······▄█▄·▄ beats 2/5	medium · avg 39%	▇ 1.8%	at its level

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #57 of 393 * · 3.8% of its score · 2 tests

Beats GPT-6 Luna, Grok 4.7, GPT-5.2 here, though they rank higher overall. Trails Gemma 4 31B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1445.6 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 308/379	avg 1268	▇ 2.9%	blunt
IFEval prompt-level strict	92.4% ✎	███··█·█···█ beats 5/5	easy · avg 74%	▇ 0.9%	typical

math▁▂▃▅▆▅▄█▄▁▁▁ #119 of 398 · 26% of its score · 7 tests

Beats Claude Sonnet 4.6, MiniMax-M3, MiMo-V2.5 (310B-A15B) here, though they rank higher overall. Trails Olmo 3.1 32B Think, Gemini 2.5 Flash, Qwen3-VL 32B Thinking here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
FrontierMath tiers 1–3 v2	24.6% ✎	▂··▁▁▃▄▄█▅▃▃ beats 16/82	very hard · avg 1.2%	▇▇▇▇ 11%	sharp · at its level
AIME 2025	89.2% ✎	▁·▁▁▁▂▅▆█▇▅▁ beats 113/133	medium · avg 51%	▇ 3.4%	sharp
AIME 2024	90.7% ✎	▁▁▁▂▁▃▁▅███▃ beats 74/85	medium · avg 60%	▇ 3.0%	sharp · maxed out
FrontierMath tier 4 v2	0.0% ✎	▃▂▁▂█▅▂▂▁▁▁▁ beats 1/63	very hard · avg 0.1%	▇ 2.8%	sharp · too hard for it
HMMT 2025 (edition unspecified)	79.3% ✎	▄··▄··▄·▄··█ beats 2/5	hard · avg 29%	▇ 2.7%	sharp
OTIS mock AIME 2024–2025	84.7% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 126/193	very hard · avg 8.5%	▇ 2.0%	sharp · noisy
MATH-500	98.0% ✎	█···██·█·█·█ beats 1/5	very easy · avg 91%	▇ 0.8%	maxed out · noisy

vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #54 of 147 · 12% of its score · 9 tests

Beats DeepSeek V4 Flash Vision (Exp), GPT-5, Claude Opus 4.5 here, though they rank higher overall. Trails Qwen3.6 Flash, Qwen3.5 35B-A3B, Gemma 4 31B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
CharXiv (reasoning questions) lab	69.6% ✎	▃·▃·▃▃▃·▃▃█▅ beats 2/11	medium · avg 40%	▇▇ 4.2%	at its level
Arena vision (style controlled)	1247.5 ✎	▁▁▂▂▂▂▄▅▅▆██ beats 101/145	avg 1068	▇ 2.0%	blunt
ChartArena chart parsing (English, mAP_high average)	47.7% ✎	▁·▁·▃▃▁▇▃█·▄ beats 19/22	hard · avg 27%	▇ 1.6%	blunt · at its level
CharXiv descriptive questions	93.4% ✎	▂▄▄▅▇▄█▇▇▃▁▂ beats 55/58	easy · avg 70%	▇ 1.2%	typical
BEAR-Bench (bilingual RU/EN enterprise and academic document images, overall accuracy)	67.4% ✎	▃▃·▅·▃·█·█▅▅ beats 12/14	hard · avg 24%	▇ 1.1%	at its level
Arena document (style controlled)	1430.6 ✎	▁▁▃▃▄▃▇█▅▇▄▅ beats 4/37	avg 1335	▇ 1.0%	blunt
Molmo2-VideoCount accuracy (val)	35.8% ✎	▃··▃·▃····█▃ beats 4/6	hard · avg 22%	▇ 0.3%	blunt · at its level
Ai2 Molmo2 human preference Elo (video/image QA, 2026)	1096.0 ✎	▄··▄··▄··▄█▄ beats 6/6	avg 975	▇ 0.2%	blunt

1 more test

test	result	vs the others in its run	difficulty	counts	why
Molmo2-CapTest dense video captioning F1	42.1% ✎	█···█·████·█ beats 2/6	hard · avg 32%	▇ 0.1%	blunt · at its level

web search▂▁▂▆▄█▃▃▁▁·▁ #59 of 80 * · 1.2% of its score · 1 test

Beats DeepSeek V4 Flash (Preview, Apr 2026), GPT-5.4 mini, Qwen3.5 397B-A17B here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), Llama 4 Scout (109B-A17B) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Arena search systems (style controlled)	1142.1 ✎	▁·▁▁▃▂▂█▅▂▁▁ beats 3/30	avg 1055	▇ 1.2%	blunt

other tests17 not in a task · 27% of its score

test	result	vs the others in its run	difficulty	counts	why
GPQA Diamond +1	85.3% ✎	▁▁▁▁▂▄▆▄▅▇█▅ beats 146/208	hard · avg 43%	▇▇▇ 8.6%	sharp · several runs
KCL-Essay vanilla	60.5% ✎	▁···▁▃▂█▂▃▃▃ beats 23/23	hard · avg 18%	▇ 2.2%	at its level
MMLU-Pro	86.7% ✎	▄·▄···█·▄··▄ beats 5/5	medium · avg 56%	▇ 2.2%	typical
MMLU-Redux	90.1% ✎	████··█····█ beats 2/5	medium · avg 73%	▇ 1.9%	typical
ZebraLogic	92.4% ✎	██··█·██···█ beats 1/5	easy · avg 75%	▇ 1.5%	maxed out
VitaBench	24.3% ✎	█····██·██·█ beats 2/5	very hard · avg 7.8%	▇ 1.4%	at its level
OJBench	41.6% ✎	▄······▄▄·▄█ beats 5/5	very hard · avg 12%	▇ 1.4%	at its level
KCL-MCQA vanilla	66.1% ✎	▂···█▂▆▆██▆▃ beats 23/23	very hard · avg 30%	▇ 1.2%	at its level

9 more tests

test	result	vs the others in its run	difficulty	counts	why
Chess Puzzles (Epoch)	20.0% ✎	█▁▁▁▁▂▁▄▂▂▁▁ beats 80/135	very hard · avg 5.9%	▇ 1.1%	sharp · noisy
RobloxQA v2.0 multiple choice (Torpedo Software)	92.8% ✎	▁▁▃▁█▄▃▃▆▂▂▂ beats 24/49	easy · avg 78%	▇ 1.0%	typical
LongSocialBench (full-context, 4-option MCQ over discussion threads)	51.5% ✎	▆▂·▄█·▂▂·▄▄▂ beats 13/16	very hard · avg 36%	▇ 1.0%	blunt · at its level
Icelandic LLM Leaderboard average (Miðeind, Sept 2026 suite)	52.2% ✎	▁▁▃▂▂▄▁▃▄▄▄█ beats 21/38	very hard · avg 14%	▇ 0.9%	at its level
BeyondAIME	63.0% ✎	█··▄····▄··█ beats 0/5	medium · avg 41%	▇ 0.9%	at its level
FinSheet-Bench accuracy (501 questions, 24 spreadsheets)	78.8% ✎	█···██·····█ beats 0/3	medium · avg 52%	▇ 0.8%	blunt
DiagnosticIQ industrial maintenance MCQ (macro accuracy, 4 options)	57.6% ✎	▁··▃▁▁▁▅█▅▁▅ beats 6/24	medium · avg 52%	▇ 0.6%	blunt · at its level
miniF2F (pass@1)	13.9% ✎	██······▄··▄ beats 2/5	very hard · avg 13%	▇ 0.1%	blunt · at its level · noisy
Vectara hallucination leaderboard (HHEM-2.3, factual consistency rate)	93.0% ✎	▁▂▁▂▅█▂▃▄▃▁▁ beats 69/98	very easy · avg 90%	▇ 0.0%	blunt

$ where to run it

Through an API

api · $3.44/M · $1.25 in / $10 out per 1M tokens · 1M context · ai.google.dev ↗ · price sourceStandard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ.

$ how the score was worked out · Meituan LongCat-Flash-Thinkin… 31%, Epoch AI — internal evaluatio… 28%, LMArena leaderboard dataset 12% and 19 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
Meituan LongCat-Flash-Thinking model card (evaluation table)	18	▇▇▇▇▇▇▇▇▇ 31%	MIT
Epoch AI — internal evaluations	6	▇▇▇▇▇▇▇▇ 28%	CC BY 4.0
LMArena leaderboard dataset	5	▇▇▇▇ 12%	CC BY 4.0
Google DeepMind Gemini model page and model cards (lab-reported scores) · lab-reported	1	▇ 4%	Factual vendor-reported score (not licensed content)
Roblox LLM Leaderboard (Torpedo Software): RobloxQA v2.0 and Luau Leetcode	2	▇ 4%	MIT
Korean Canonical Legal Benchmark: Toward Knowledge-Independent Evaluation of LLMs’ Legal Reasoning Capabilities	2	▇ 3%	CC BY 4.0
ContractBench: Can LLM Agents Preserve Observation Contracts? (arXiv:2605.17281v2)	1	▇ 3%	CC BY 4.0
Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces — benchmark authors' own agent-model trials	1	▇ 3%	CC BY 4.0
ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)	1	▇ 2%	CC BY 4.0
τ²/τ³-bench leaderboard: Sierra's own evaluation runs	1	▇ 1%	MIT
Baidu - ERNIE 5.0 Technical Report (arXiv:2602.04705v1)	1	▇ 1%	CC BY 4.0
BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)	1	▇ 1%	CC BY 4.0
LongSocialBench: Do Long-Context LLMs Understand Online Discussion Threads? (arXiv:2610.04118v1)	1	▇ 1%	CC BY 4.0
Miðeind Icelandic LLM Leaderboard	1	▇ 1%	CC BY 4.0
EEGAgentBench: Benchmarking LLM Agents on Short- and Long-Horizon EEG Analysis (arXiv:2609.31632v1)	1	▇ 1%	CC BY 4.0
FinSheet-Bench: From Simple Lookups to Complex Reasoning, Where LLMs Break on Financial Spreadsheets — authors' own model runs	1	▇ 1%	CC BY 4.0
Berkeley Function Calling Leaderboard result archive	1	▇ 1%	Apache-2.0
Molmo2: Open Weights and Data for Vision-Language Models with Video Understanding and Grounding (Ai2, arXiv 2601.10611v4)	3	▇ 1%	CC BY 4.0
Galileo Agent Leaderboard (v1 and v2 results)	1	▇ 1%	Apache-2.0
DiagnosticIQ: A Benchmark for LLM-Based Industrial Maintenance Action Recommendation from Symbolic Rules (arXiv:2605.08614v1)	1	▇ 1%	CC BY 4.0
Amazon AGI — τ²-Bench-Verified	1	▇ 1%	MIT
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	1	▇ 0%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 67.2