whatmodel.sh / Gemini 2.5 Flash

Gemini 2.5 Flash

+ contribute

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

Catalogue identity notes

Release month: Gemini API changelog, June 17, 2025: 'Released gemini-2.5-flash, our first stable 2.5 Flash model.'

Model information sources

Release date ↗

61.5±0.7#183 of 719 LLMsfrom 42 benchmark results across 16 sources

$ how it compares

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

#	model	score	where	RAM	released
214	Apriel 1.6 15B Thinker ServiceNow	62±2	local	~10GB	2025-11
215	Gemini 2.5 Flash-Lite Preview (06-17) Google	62±2	comparison only		2025-06
216	Gemini 2.5 Flash Google	62±1	api $0.85/M		2025-06
217	MiniMax-M1 (40k) MiniMax	61±1	comparison only		2025-06
218	Grok 4 Fast SpaceXAI (xAI)	61±2	comparison only		2025-09
	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▁▁▃▆█▆▅▅▅▄▃▁ #180 of 300 · 24% of its score · 10 tests

Beats o1 (2024-12-17), o3-mini, DeepSeek R1 0528 here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Mellum2 12B-A2.5B Instruct here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
ContractBench success rate (33 observation-contract tool tasks)	19.2% ✎	▆··▆▄▂▄▂▆▆█▄ beats 5/23	very hard · avg 14%	▇▇ 4.2%	sharp · at its level
SWE-bench Verified	40.6% ✎	▁▁▁▁▁▁▃▄█▆▁▁ beats 27/131	hard · avg 18%	▇▇ 4.0%	sharp · at its level
Terminal-Bench 2.0 with Terminus 2 agent (benchmark authors' runs)	16.9% ✎	▃··▃·▃▃██▅▃▅ beats 2/14	very hard · avg 5.8%	▇ 3.1%	sharp
τ²-bench Retail	64.8% ✎	▅···█·▃····▃ beats 0/6	medium · avg 41%	▇ 2.9%	at its level
BFCL V4 (December 2025 archive)	56.2% ✎	▁▁▁▂██▅▅▆▂▂▂ beats 67/77	hard · avg 32%	▇ 2.3%	at its level
τ²-bench Telecom	16.5% ✎	██··███···██ beats 0/6	very hard · avg 9.6%	▇ 2.2%	sharp · at its level · noisy
τ²-bench Airline	41.5% ✎	▄·▄··▄···█▄▄ beats 1/6	hard · avg 26%	▇ 1.7%	at its level
Terminal-Bench (pre-2.0 release)	12.3% ✎	▄·▄···▄▄···█ beats 0/5	very hard · avg 12%	▇ 1.1%	typical

2 more tests

test	result	vs the others in its run	difficulty	counts	why
τ²-Bench-Verified (GPT-5.1 user)	41.6% ✎	▃··▄···▂▄▄█▄ beats 0/13	medium · avg 45%	▇ 1.0%	at its level
Galileo Agent Leaderboard v2 (action completion)	38.0% ✎	▅▃·▃▃▅▅█▃█▅▅ beats 8/19	hard · avg 27%	▇ 1.0%	at its level

coding▁▁▁▃▆██▇▆▄▃▁ #198 of 458 · 14% of its score · 4 tests

Trails QwQ-32B-Preview, Ministral 3 3B Reasoning 2512, DeepSeek-Coder-V2 Instruct (0724) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
SWE-bench Verified	40.6% ✎	▁▁▁▁▁▁▃▄█▆▁▁ beats 27/131	hard · avg 18%	▇▇ 4.0%	sharp · at its level
LiveCodeBench (version unspecified)	39.6% ✎	█·······▄▄▄▄ beats 1/5	hard · avg 21%	▇ 3.7%	at its level
HumanEval+	87.8% ✎	▄▄▄·····▄▄·█ beats 1/6	easy · avg 67%	▇ 3.2%	sharp
MBPP+	76.2% ✎	▃·····▃█▃··▃ beats 0/6	medium · avg 61%	▇ 2.8%	typical

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #103 of 393 * · 6.8% of its score · 2 tests

Beats LongCat 2.0 (1.6T-A48B), Inkling Small (266B-A12B), GPT-5.4 nano here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1409.4 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 255/379	avg 1268	▇▇ 4.6%	blunt
IFEval (aggregation unspecified)	83.9% ✎	█··█·█··████ beats 0/6	easy · avg 75%	▇ 2.2%	typical

math▁▂▃▅▆▅▄█▄▁▁▁ #85 of 398 · 6.5% of its score · 4 tests

Beats Claude Sonnet 4.6, Claude Opus 4.5, MiniMax-M3 here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
AIME 2025	67.3% ✎	▄▄···▄▄··▄·█ beats 5/6	hard · avg 19%	▇ 2.3%	sharp · at its level · noisy
MathNet-Solve-Test micro-average accuracy (6,400 olympiad problems)	41.1% ✎	▃▃···█▃·▃·▃▃ beats 3/8	hard · avg 15%	▇ 1.5%	at its level
AIME 2024	79.7% ✎	▁▁▁▂▁▃▁▅███▃ beats 52/85	hard · avg 32%	▇ 1.4%	sharp · at its level · noisy
MATH-500	98.4% ✎	▄·▄··█··▄·▄▄ beats 5/6	very easy · avg 86%	▇ 1.3%	maxed out · noisy

vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #76 of 147 · 16% of its score · 7 tests

Beats DeepSeek V4 Flash Vision (Exp), GPT-5.4 nano, GPT-5 mini here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
CharXiv (reasoning questions) lab +1	63.7% ✎	▃·▃·▃▃▃·▃▃█▅ beats 1/11	medium · avg 40%	▇▇▇▇ 9.4%	at its level · several runs
Arena vision (style controlled)	1215.1 ✎	▁▁▂▂▂▂▄▅▅▆██ beats 80/145	avg 1068	▇ 3.2%	blunt
BEAR-Bench (bilingual RU/EN enterprise and academic document images, overall accuracy)	49.0% ✎	▃▃·▅·▃·█·█▅▅ beats 6/14	hard · avg 24%	▇ 1.7%	at its level
CharXiv descriptive questions	83.6% ✎	▄····▄···█▄▄ beats 1/5	easy · avg 68%	▇ 1.0%	noisy
Molmo2-VideoCount accuracy (val)	31.9% ✎	▃··▃·▃····█▃ beats 2/6	hard · avg 22%	▇ 0.4%	blunt · at its level
Ai2 Molmo2 human preference Elo (video/image QA, 2026)	1084.0 ✎	▄··▄··▄··▄█▄ beats 5/6	avg 975	▇ 0.3%	blunt
Molmo2-CapTest dense video captioning F1	46.0% ✎	█···█·████·█ beats 4/6	hard · avg 32%	▇ 0.2%	blunt · at its level

other tests15 not in a task · 37% of its score

test	result	vs the others in its run	difficulty	counts	why
MMLU	86.3% ✎	▄·····█▄▄▄·▄ beats 0/6	medium · avg 71%	▇▇ 6.1%	typical
GPQA Diamond	80.3% ✎	█·█·█··██··█ beats 5/5	hard · avg 42%	▇▇ 5.2%	sharp · at its level
MMLU-Pro	82.0% ✎	█▄·▄···▄··▄▄ beats 1/6	medium · avg 59%	▇▇ 4.6%	typical
KCL-Essay vanilla	53.7% ✎	▁···▁▃▂█▂▃▃▂ beats 21/23	hard · avg 18%	▇ 3.3%	at its level
CMMLU	78.3% ✎	█····▄··▄▄▄▄ beats 1/6	medium · avg 67%	▇ 2.8%	typical
C-Eval	78.8% ✎	█····▄·▄·▄▄▄ beats 0/6	medium · avg 68%	▇ 2.7%	typical
BRIDGE clinical text, zero-shot average (87 tasks)	44.8% ✎	▁▁▁▁▁▂▂▃▄▆█▃ beats 94/97	hard · avg 32%	▇ 2.6%	blunt · at its level
DROP	45.0% ✎	▄···▄·▄█·▄·▄ beats 0/6	medium · avg 41%	▇ 2.1%	at its level · noisy

7 more tests

test	result	vs the others in its run	difficulty	counts	why
COLLIE (constrained text generation)	48.6% ✎	▄···▄█▄····█ beats 1/6	hard · avg 30%	▇ 1.9%	at its level
VitaBench	8.0% ✎	█·······▄█·█ beats 0/6	very hard · avg 6.4%	▇ 1.7%	typical
KCL-MCQA vanilla	51.8% ✎	▂···█▂▆▆██▆▂ beats 19/23	very hard · avg 30%	▇ 1.6%	at its level
ZebraLogic	51.8% ✎	█···█··███·█ beats 0/5	medium · avg 53%	▇ 0.9%	noisy
Open PT-LLM Leaderboard average (9 Portuguese tasks)	85.6% ✎	▃▁▁▁▁▁▁▁▂█▅▃ beats 275/276	medium · avg 73%	▇ 0.7%	noisy
BeyondAIME	44.2% ✎	█····█·▄▄··▄ beats 5/6	hard · avg 25%	▇ 0.6%	at its level
Vectara hallucination leaderboard (HHEM-2.3, factual consistency rate)	92.2% ✎	▁▂▁▂▅█▃▃▄▃▁▁ beats 63/98	very easy · avg 90%	▇ 0.0%	blunt

$ where to run it

Through an API

api · $0.85/M · $0.3 in / $2.5 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-Chat mo… 58%, LMArena leaderboard dataset 8%, Google DeepMind Gemini model… 8% and 13 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-Chat model card (evaluation table)	22	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 58%	MIT
LMArena leaderboard dataset	2	▇▇ 8%	CC BY 4.0
Google DeepMind Gemini model page and model cards (lab-reported scores) · lab-reported	1	▇▇ 8%	Factual vendor-reported score (not licensed content)
Korean Canonical Legal Benchmark: Toward Knowledge-Independent Evaluation of LLMs’ Legal Reasoning Capabilities	2	▇ 5%	CC BY 4.0
ContractBench: Can LLM Agents Preserve Observation Contracts? (arXiv:2605.17281v2)	1	▇ 4%	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
ServiceNow-AI/Apriel-1.6-15b-Thinker model card	2	▇ 3%	MIT
BRIDGE Medical Leaderboard (Harvard Medical School / Brigham and Women's YLab)	1	▇ 3%	MIT
Berkeley Function Calling Leaderboard result archive	1	▇ 2%	Apache-2.0
BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)	1	▇ 2%	CC BY 4.0
MathNet: a Global Multimodal Benchmark for Mathematical Reasoning and Retrieval — authors' own model runs	1	▇ 2%	CC BY 4.0
Amazon AGI — τ²-Bench-Verified	1	▇ 1%	MIT
Galileo Agent Leaderboard (v1 and v2 results)	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
Open Portuguese LLM Leaderboard (archived)	1	▇ 1%	Apache-2.0
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	1	▇ 0%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 61.5