[whatmodel.sh](../../README.md) / Gemini 3 Flash (Preview)
# Gemini 3 Flash (Preview)
Google · released 2025-12 · closed weights
**score 71.0** ±0.6 · #69 of 732 · from 30 benchmark results across 14 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Gemini 3 Flash (Preview) among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆█▄▅●▄▄▂▁ #82 of 313 16% 6
coding ▁▁▁▃▅▇█▇▅●▃▁ #73 of 469 16% 3
chat & writing ▁▁▁▃▆▄▆█▆▇●▃ #32 of 402* 3.9% 1
math ▁▁▃▅█▆▄█●▂▁▁ #55 of 431 37% 6
vision & documents ▁▁▂▂▁▃▄▂▆▇●▄ #29 of 153 10% 5
web search ▁▁▁▅▄●▃▄·▁·▁ #41 of 84* 1.4% 1
agent tasks · #82 of 313 · 16% of its score
Beats [GPT-5](gpt-5.md) here, though they rank higher overall. Trails [MiMo-V2.5 (310B-A15B)](mimo-v2-5.md), [Qwen3.6 27B](qwen3-6-27b.md), [Muse Glimmer 30B](muse-glimmer-30b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
SWE-bench Verified 75.4% ▁··▁·▁▂▁▄●▄▁ beats 19/29 hard · avg 19% 7.5% sharp · at its level
τ³-bench banking_knowledge (v1.0.… 27.3% ▃▃▂▃·▆●▂█▅▅▂ beats 12/25 very hard · avg 4.3% 2.4% at its level
Vibe Code Bench (test split, 50 w… 20.2% ▃··▃·▃█●▃▃▃█ beats 7/14 very hard · avg 2.8% 1.7% sharp · at its level · noisy
τ³-bench airline (tau2-bench ≥ v1… 82.5% ▃······▃·▃●▃ beats 4/6 medium · avg 51% 1.5% typical
τ³-bench retail (tau2-bench ≥ v1.… 76.8% ▄▄·●··▄·▄··▄ beats 3/6 medium · avg 48% 1.4% typical
τ³-bench telecom (tau2-bench ≥ v1… 91.2% █·▄●▄······▄ beats 4/6 medium · avg 64% 1.1% typical
coding · #73 of 469 · 16% of its score
Beats [DeepSeek V4 Flash Vision (Exp)](deepseek-v4-flash-vision-exp.md), [GPT-5.2](gpt-5-2.md), [GPT-5](gpt-5.md) here, though they rank higher overall. Trails [Llama 3.1 Nemotron Ultra 253B v1](llama-3-1-nemotron-ultra-253b.md), [MiMo-V2.5-Pro (1T-A42B)](mimo-v2-5-pro.md), [Qwen3.6 Plus](qwen3-6-plus.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
SWE-bench Verified 75.4% ▁··▁·▁▂▁▄●▄▁ beats 19/29 hard · avg 19% 7.5% sharp · at its level
Arena WebDev (Code Arena) 1440.1 ▁▁▃▂▅●▃▆▅▅▁▂ beats 61/129 avg 981 6.9% sharp
Vibe Code Bench (test split, 50 w… 20.2% ▃··▃·▃█●▃▃▃█ beats 7/14 very hard · avg 2.8% 1.7% sharp · at its level · noisy
chat & writing · #32 of 402 · 3.9% of its score
Beats [GPT-6 Sol](gpt-6-sol.md), [GPT-5.6 Terra](gpt-5-6-terra.md), [MiMo-V2.6-Flash (309B-A15B)](mimo-v2-6-flash.md) here, though they rank higher overall.
test result vs the others in its run difficulty counts why
LMArena (text) 1472.7 ▁▁▁▂▃▃▃▆▆▇●▃ beats 349/385 avg 1260 3.9% blunt
math · #55 of 431 · 37% of its score
Beats [Qwen3.6 Max Preview](qwen3-6-max-preview.md), [Claude Sonnet 4.6](claude-sonnet-4-6.md) here, though they rank higher overall. Trails [A.X K2 (688B-A33B)](ax-k2.md), [DeepSeek V4 Flash (Preview, Apr 2026)](deepseek-v4-flash-preview.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
FrontierMath tiers 1–3 v2 51.2% ▂··▁▁▃▄▄●▆▄▃ beats 35/86 very hard · avg 1.2% 19% sharp · at its level
FrontierMath tier 4 v2 17.1% ▁▂▁▂●▅▂▃▂▁▁▁ beats 17/67 very hard · avg 0.1% 11% sharp
MathArena ArXivMath January 2026 … 58.2% ▂··▄▂▄█●▆▄▄█ beats 12/23 hard · avg 23% 3.1% at its level
MathArena ArXivMath December 2025… 43.4% ▃·▃▃█●·▄·▁▁▁ beats 12/18 hard · avg 19% 2.3% at its level
OTIS mock AIME 2024–2025 95.6% ▃▃▄▃▃▃▇▇█●▂▅ beats 161/197 very hard · avg 7.2% 1.0% sharp · noisy
FormalProofBench (200 graduate-le… 15.0% ▃▅·▃··▅●█··▃ beats 6/10 very hard · avg 3.1% 0.7% noisy
vision & documents · #29 of 153 · 10% of its score
Beats [Claude Sonnet 5.5](claude-sonnet-5-5.md), [GPT-6 Sol](gpt-6-sol.md), [MiMo-V2.6-Pro](mimo-v2-6-pro.md) here, though they rank higher overall. Trails [Qwen3.6 Plus](qwen3-6-plus.md), [Qwen3.7 Plus](qwen3-7-plus.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
CharXiv (reasoning questions) (ma… 80.3% ▃·▃·▃▃▃·●▃█▅ beats 5/11 medium · avg 38% 4.6% typical
Arena vision (style controlled) 1272.5 ▁▁▂▂▂▂▃▄▅▅●▇ beats 125/152 avg 1061 2.6% blunt
Arena document (style controlled) 1439.2 ▁▁▃▃●▃▇█▅▇▄▅ beats 7/37 avg 1331 1.3% blunt
Diagram-MMU diagram question answ… 84.1% ▄·····▄▄·●·█ beats 4/6 medium · avg 56% 0.5% blunt
SpaceCast-Bench predictive spatia… 56.7% ▃·▅▃▅▃··▅··● beats 10/11 very hard · avg 33% 0.4% at its level · noisy
web search · #41 of 84 · 1.4% of its score
Beats [Gemini 3.1 Pro (Preview)](gemini-3-1-pro.md), [Gemini 3.5 Flash](gemini-3-5-flash.md), [Qwen3.7 Max](qwen3-7-max.md) here, though they rank higher overall. Trails [GPT-4o mini (2024-07-18)](gpt-4o-mini-2024-07-18.md), [Mistral Small 3.1 24B (2503)](mistral-small-3-1.md), [Llama 4 Scout (109B-A17B)](llama-4-scout.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena search systems (style contr… 1188.7 ▁·▁▁▂▂▂●▆▃▂▂ beats 13/36 avg 1069 1.4% blunt
other tests (10)
test result vs the others in its run difficulty counts why
Mystery Game Puzzles (Epoch) 20.0% █··▂▃●▄█▂▄▂▁ beats 34/77 very hard · avg 9.4% 3.9% sharp · noisy
WebCompass (overall) 65.4% ▅·▃▃▃···▃·▃● beats 7/9 hard · avg 18% 3.8% at its level
GPQA Diamond 89.4% ▁▁▁▁▂▄▆▄▅▇●▅ beats 169/211 hard · avg 39% 3.7% sharp
ARC-AGI-2 public evaluation set (… 34.0% █··▂▅▁▃▁●·▁▃ beats 18/24 very hard · avg 0.3% 3.7% sharp · at its level · noisy
KINA (Knowledge Index of Noah's A… 43.9% ▁▁▂▁▃▄█▆▄▅●▃ beats 36/39 very hard · avg 21% 3.1% at its level
Chess Puzzles (Epoch) 40.0% █▁▁▁▁▂▃▃●▂▁▁ beats 118/137 very hard · avg 5.8% 2.8% sharp · noisy
ARC-AGI-1 public evaluation set (… 31.9% ▁·▇▃●·█▇▁·▇▃ beats 8/27 very hard · avg 4.0% 2.2% sharp · noisy
SimpleQA Verified 66.8% ▁▁▃▂▃▃▆▆█▄●▅ beats 72/81 very hard · avg 8.0% 1.5% at its level
SCHEDBench subset, solver-mediate… 12.0% ▄·●·····▄··█ beats 1/4 very hard · avg 5.4% 0.9% at its level · noisy
Vectara hallucination leaderboard… 86.5% ▁▂▁●▅█▂▃▄▃▁▁ beats 16/98 very easy · avg 90% 0.0% blunt
source results share of the score
Epoch AI 8 50% ▇▇▇▇▇▇
LMArena leaderboard dataset 5 16% ▇▇
τ²/τ³-bench leaderboard: Sierra's own e… 4 6% ▇
ARC Prize Foundation model_baseline out… 2 6% ▇
MathArena ArXivMath / BrokenArXiv / ArX… 2 5% ▇
Google DeepMind Gemini model page and m… 1 5% ▇
WebCompass: Towards Multimodal Web Codi… 1 4% ▇
Knowledge Index of Noah's Ark 1 3% ▇
Vibe Code Bench 1 2% ▇
Improving Natural-Language Combinatoria… 1 1% ▇
FormalProofBench 1 1% ▇
Diagram-MMU: A Multi-Modal Benchmark fo… 1 1% ▇
SpaceCast-Bench: Evaluating Predictive … 1 0% ▇
Vectara hallucination leaderboard 1 0% ▇
prior: pull toward the middle 0% ▇
weighted average = 71.0
Licences: Epoch AI ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · τ²/τ³-bench leaderboard: Sierra's own evaluation runs ([MIT](https://opensource.org/license/mit)) · ARC Prize Foundation model\_baseline outputs on ARC-AGI-1/2 public evaluation sets ([MIT](../../licenses/arcprize-model-baseline-outputs-MIT.txt)) · MathArena ArXivMath / BrokenArXiv / ArXivLean model-output datasets ([CC BY-SA 4.0](../../licenses/matharena-sharealike-NOTICE.txt)) · WebCompass: Towards Multimodal Web Coding Evaluation for Code Language Models ([CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/)) · Knowledge Index of Noah's Ark ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Vibe Code Bench ([CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)) · Improving Natural-Language Combinatorial-Optimization Accuracy in Resource-Constrained Language Models via Formal Abstractions ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · FormalProofBench ([CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)) · Diagram-MMU: A Multi-Modal Benchmark for Scientific Diagrams ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Vectara hallucination leaderboard ([Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0))
## How to run it
### Through an API
api (Google) · $1.13/M blended · $0.5 in / $3 out per 1M tokens · [ai.google.dev](https://ai.google.dev/gemini-api/docs/models/gemini-3-flash-preview)
[price source](https://ai.google.dev/gemini-api/docs/pricing): USD per 1M tokens, Gemini Developer API Standard paid tier, gemini-3-flash-preview (Gemini 3 Flash Preview, which Google calls 'our legacy Flash model'; the deprecations table lists 'No shutdown date announced'): input $0.50 (text/image/video), output $3.00 (including thinking tokens). Audio input ($1.00), context caching, Batch and Flex ($0.25 in / $1.50 out) and the free tier are excluded. Currency: USD. Google Gemini API official pricing page, observed 2026-09-26. · [context source](https://ai.google.dev/gemini-api/docs/models/gemini-3-flash-preview): Gemini Developer API model page, gemini-3-flash-preview: input token limit 1,048,576 (the 65,536-token output limit is separate). Observed 2026-09-26.
[interactive version](https://whatmodel.sh/model/gemini-3-flash-preview) · [JSON](https://whatmodel.sh/api?format=json&all=1)