whatmodel.sh / Gemini 3.8 Flash

Gemini 3.8 Flash

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Gemini 3.8 Flash is a closed-weights language model from Google, released September 2026.

77.2±0.6#23 of 719 LLMsfrom 33 benchmark results across 11 sources

$ how it compares

The LLMs ranked either side of Gemini 3.8 Flash, out of 828 with a score, then more from Google.

#modelscorewhereRAMreleased
21DeepSeek V4.1 Flash DeepSeek78±1api $0.26/M +local~537GB2026-09
22Claude Haiku 5.5 Anthropic77±1api $0.20/M2026-10
23Gemini 3.8 Flash Google77±1api $1.50/M2026-09
24Muse Spark 1.3 Meta77±1api $2.00/M2026-09
25Kimi K3 Moonshot AI77±1api $6.00/M +local~884GB2026-07
more from Google
9Gemini 4 Argon (limited preview) Google82±1limited preview2026-09
32Gemini 3.7 Flash Google76±1api $1.50/M2026-08
39Gemini 3.1 Pro (Preview) Google75±1api $4.50/M2026-02

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #27 of 300 · 20% of its score · 9 tests

Beats GPT-5.4 here, though they rank higher overall. Trails GLM-5.3-Flash, Grok 4.7, Qwen3.8 Max (0902) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena Agent (October 2026 aggregate IPS)0.0297 IPS ✎▁▁▁▄▂▃▇█▂▂▁▃ beats 29/49avg -0.312 IPS▇▇ 5.8%sharp
Terminal-Bench 4.019.1% ✎█▅▅▅▅·▅▃▃▅▅▅ beats 6/20very hard · avg 0.4%▇ 3.4%sharp · at its level
Terminal-Bench 2.1 lab +190.8% ✎▁·▁·▁▁▄▃▄█▃▇ beats 56/56hard · avg 23%▇ 3.3%small run · several runs
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@136.7% ✎▃··▅··▃█▃▅▃▅ beats 3/12very hard · avg 3.1%▇ 2.2%sharp · at its level
SWE-bench Pro (split unspecified) lab61.6% ✎▃·▁▁▁▂▂▃█▇▃▁ beats 42/55very hard · avg 6.4%▇ 1.9%sharp · at its level · small run
Video2World V2WScore (coding agents reconstruct interactive simulated worlds from videos, OpenCode scaffold)30.1% ✎█··▄█·▄····▄ beats 2/6very hard · avg 6.6%▇ 1.4%at its level
DeepSWE v1.1 (resolved) lab73.7% ✎▂····▁▁▁▆██▆ beats 24/28hard · avg 23%▇ 1.0%at its level · small run
KlinikeBench strict pass@1 (333 interactive clinical encounters, Terminus-2)18.9% ✎▁▁·▁▁▁▁▅▃█▅█ beats 18/27very hard · avg 2.4%▇ 0.9%typical
1 more test
testresultvs the others in its rundifficultycountswhy
τ³-bench Banking lab38.1% ✎▂▂·▅█▂▂▃▂▅▂▃ beats 20/20very hard · avg 6.9%▇ 0.4%blunt · small run
coding▁▁▁▃▆██▇▆▄▃▁ #29 of 458 · 12% of its score · 5 tests

Beats GPT-5.5, GPT-5.6 Terra, Claude Opus 4.8 here, though they rank higher overall. Trails GLM-5.2, GLM-5.3-Flash, Grok 4.7 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena WebDev (Code Arena)1582.7 ✎▁▂▄▃▅█▄▇▆▄▁▁ beats 92/119avg 981▇▇ 5.1%sharp
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@136.7% ✎▃··▅··▃█▃▅▃▅ beats 3/12very hard · avg 3.1%▇ 2.2%sharp · at its level
SWE-bench Pro (split unspecified) lab61.6% ✎▃·▁▁▁▂▂▃█▇▃▁ beats 42/55very hard · avg 6.4%▇ 1.9%sharp · at its level · small run
Luau Leetcode fully solved (Torpedo Software)84.6% ✎▁▁▁▁·▁▂▃▅▃█▇ beats 44/49hard · avg 18%▇ 1.8%typical
DeepSWE v1.1 (resolved) lab73.7% ✎▂····▁▁▁▆██▆ beats 24/28hard · avg 23%▇ 1.0%at its level · small run
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #7 of 393 * · 2.9% of its score · 1 test

Beats GPT-6 Astra, GPT-6.1 Sol, Claude Sonnet 5.5 here, though they rank higher overall. Trails Claude Opus 4.6, Claude Opus 4.7 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LMArena (text)1494.8 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 373/379avg 1268▇ 2.9%blunt
math▁▂▃▅▆▅▄█▄▁▁▁ #30 of 398 · 47% of its score · 10 tests

Beats DeepSeek V4.1 Flash here, though they rank higher overall. Trails Qwen3.7 Max, Gemini 3.1 Pro (Preview), Qwen3.8 Max (0902) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
FrontierMath tier 4 v222.0% ✎▁▂▁▂█▅▂▂▁▁▁▁ beats 24/63very hard · avg 0.1%▇▇▇▇▇ 13%sharp · at its level
FrontierMath tiers 1–3 v268.4% ✎▂··▁▁▃▄▄█▅▃▃ beats 57/82very hard · avg 1.2%▇▇▇▇▇ 12%sharp · at its level
MathArena ArXivMath June 2026 (48 problems)67.4% ✎▂···▄·██▆▆▄▂ beats 10/19very hard · avg 4.6%▇▇▇ 6.9%sharp · at its level
MathArena ArXivMath April 2026 (40 problems)64.2% ✎▁·····▅██▄·▁ beats 16/20very hard · avg 7.2%▇▇ 6.2%sharp · at its level
MathArena BrokenArXiv April 2026 (61 problems)33.6% ✎▂▂▆█▃▃▂▂▂··▂ beats 11/18very hard · avg 0.7%▇ 2.8%sharp · at its level · noisy
MathArena ArXivMath May 2026 (40 problems)58.3% ✎▁··▁▄█▄▁·▃·▁ beats 10/17very hard · avg 10%▇ 2.2%at its level
MathArena ArXivLean June 2026 (46 problems)19.6% ✎▄····▄▄███·▄ beats 4/9very hard · avg 0.7%▇ 2.0%sharp
MathArena ArXivMath August 2026, with tools (57 problems)40.4% ✎█▄·▄·······█ beats 1/5very hard · avg 4.8%▇ 1.5%sharp · at its level
2 more tests
testresultvs the others in its rundifficultycountswhy
MathArena BrokenArXiv August 2026, with tools (56 problems)27.4% ✎▄▄·····▄·▄·█ beats 1/5very hard · avg 5.5%▇ 0.6%at its level · noisy
OTIS mock AIME 2024–202598.9% ✎▃▃▄▃▃▃▇▇█▃▂▅ beats 180/193very hard · avg 8.5%▇ 0.4%sharp · maxed out · noisy
vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #7 of 147 * · 4.6% of its score · 2 tests

Beats GPT-6 Astra, GPT-6.1 Sol, Claude Fable 5.1 here, though they rank higher overall. Trails Claude Opus 4.7, Gemini 3.7 Flash here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
CharXiv (reasoning questions) lab86.2% ✎▃·▃·▃▃▃·▃▃█▅ beats 11/11medium · avg 40%▇ 2.6%typical
Arena vision (style controlled)1290.3 ✎▁▁▂▂▂▂▄▅▅▆██ beats 137/145avg 1068▇ 2.0%blunt
other tests8 not in a task · 18% of its score
testresultvs the others in its rundifficultycountswhy
Furniture Assembly (Epoch)31.7% ✎█▆▆▆▆▂▅▂▂▂·▅ beats 9/30very hard · avg 1.2%▇▇ 5.1%sharp · at its level
Mystery Game Puzzles (Epoch)47.0% ✎█··▂▃▇▄█▃▃▁▁ beats 64/74very hard · avg 9.4%▇▇ 4.6%sharp · at its level · noisy
Chess Puzzles (Epoch)61.0% ✎█▁▁▁▁▂▁▄▂▂▁▃ beats 133/135very hard · avg 5.9%▇ 2.7%sharp · at its level · noisy
GPQA Diamond95.4% ✎▁▁▁▁▂▄▆▄▅▇█▅ beats 206/208hard · avg 40%▇ 1.8%sharp
SimpleQA Verified69.7% ✎▁▁▃▂▃▃▆▆█▃▂▅ beats 72/78very hard · avg 8.0%▇ 1.2%at its level
Icelandic LLM Leaderboard average (Miðeind, Sept 2026 suite)63.5% ✎▁▁▃▂▂▄▁▃▄▄▄█ beats 28/38very hard · avg 14%▇ 1.0%at its level
LVBench lab87.1% ✎▇·█▅▅▄▄··▃▃▁ beats 27/28medium · avg 57%▇ 0.9%typical
RobloxQA v2.0 multiple choice (Torpedo Software)96.5% ✎▁▁▃▁█▄▃▃▆▂▂▂ beats 35/49easy · avg 78%▇ 0.6%maxed out

$ where to run it

Through an API

api · $1.50/M · $0.75 in / $3.75 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 · Epoch AI — internal evaluatio… 40%, MathArena ArXivMath / BrokenA… 22%, LMArena leaderboard dataset 16% and 8 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

sourceresultsshare of the scorelicence
Epoch AI — internal evaluations8▇▇▇▇▇▇▇▇▇▇▇▇ 40%CC BY 4.0
MathArena ArXivMath / BrokenArXiv / ArXivLean model-output datasets7▇▇▇▇▇▇▇ 22%CC BY-SA 4.0
LMArena leaderboard dataset4▇▇▇▇▇ 16%CC BY 4.0
Google DeepMind Gemini model page and model cards (lab-reported scores) · lab-reported4▇▇ 6%Factual vendor-reported score (not licensed content)
Google Cloud developer's guides to Gemini 3.8 Flash and 3.7 Flash · lab-reported3▇ 4%CC-BY-4.0
Terminal-Bench 4.0 official leaderboard submissions1▇ 3%Apache-2.0
Roblox LLM Leaderboard (Torpedo Software): RobloxQA v2.0 and Luau Leetcode2▇ 2%MIT
E2E-SWE: Benchmarking LLMs on Building Working Codebases from Scratch (arXiv:2609.38335v1)1▇ 2%CC BY 4.0
Video2World: Benchmarking Coding Agents for Interactive World Modeling from Embodied Videos (arXiv:2610.04432v2)1▇ 1%CC BY-SA 4.0
Miðeind Icelandic LLM Leaderboard1▇ 1%CC BY 4.0
KlinikeBench: Evaluating Language Models Beyond Diagnostic Accuracy (arXiv:2609.38480v2)1▇ 1%CC BY 4.0
prior · pull toward the middle▇ 0%
weighted average= 77.2

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json