benchgap
Google · model

Gemini 3.8 Flash benchmark scores

As of 2026-11-22, Gemini 3.8 Flash (Google) has measured scores on 65 benchmarks and estimated scores on 121 more.

BenchmarkScoreSource
AA-GPQA Diamond95.3%measured
AA-HLE47.8%measured
AA-SciCode56.6%measured
AIME2695.9%estimated ± 1.2 pp, medium confidence
Artificial Analysis Intelligence Index30.9%measured
CritPt18.3%measured
GDPval-AA46.8%measured
GPQA Diamond96.4%estimated ± 1.4 pp, high confidence
HLE57.2%estimated ± 2.2 pp, high confidence
AA-LCR81.3%measured
LiveCodeBench v690.7%estimated ± 2.2 pp, low confidence
AA-Omniscience Accuracy50.5%measured
SciCode55.8%estimated ± 2.9 pp, low confidence
Terminal-Bench 2.189.4%measured
BrowseComp91.6%estimated ± 1.9 pp, high confidence
HLE w/ tools45.5%estimated ± 3.1 pp, medium confidence
LongBench v266.8%estimated ± 5.5 pp, medium confidence
τ²-bench results93.2%estimated ± 6.5 pp, low confidence
AA Coding Index76.3%measured
AA-MMMU-Pro85.6%measured
APEX-Agents39.2%estimated ± 9.4 pp, low confidence
APEX-Agents-AA38.7%estimated ± 2.9 pp, low confidence
BioMysteryBench (human-difficult)56.5%measured
DeepSearchQA91.9%estimated ± 4.5 pp, high confidence
SWE-bench Verified81.9%estimated ± 4.4 pp, high confidence
AutomationBench39.2%estimated ± 9.5 pp, low confidence
BFCL v476.4%estimated ± 2.7 pp, medium confidence
CyberGym85.6%estimated ± 4.5 pp, high confidence
JobBench64.6%estimated ± 4.9 pp, medium confidence
SWE-bench Pro73.4%estimated ± 4.5 pp, high confidence
WideResearch84.9%estimated ± 3.3 pp, medium confidence
AA Agentic Index41.1%measured
MMLU-Pro87.8%estimated ± 4.8 pp, medium confidence
FailureSensorIQ69.0%measured
GPQA93.6%estimated ± 0.9 pp, high confidence
MMLU90.5%estimated ± 0.9 pp, medium confidence
Vals GPQA Diamond94.4%measured
Vals LiveCodeBench89.5%measured
Vals MMLU-Pro90.2%measured
MMMLU90.6%estimated ± 0.7 pp, low confidence
MMMU89.7%estimated ± 2.7 pp, medium confidence
Chess Puzzles58.0%measured
AA MMLU-Pro88.9%estimated ± 0.7 pp, medium confidence
Gert Labs67.0%measured
AA-AnalystAgent46.3%measured
AA EnterpriseOps-Gym35.4%estimated ± 7.2 pp, low confidence
AA Harvey LAB v1.092.4%estimated ± 1.2 pp, medium confidence
AA Tau3 Banking47.1%estimated ± 2.4 pp, low confidence
AA Terminal-Bench 2.187.6%measured
ApprenticeBench24.0%measured
ARC-AGI-198.5%measured
ARC-AGI-289.2%measured
BabaIsBench100.0%measured
CursorBench 3.160.3%estimated ± 3.4 pp, medium confidence
CursorBench 3.269.2%measured
EBR-bench18.9%estimated ± 9.3 pp, medium confidence
Terminal-Bench 3.021.4%estimated ± 3.0 pp, high confidence
FrontierCode 1.1 Main52.8%estimated ± 1.6 pp, medium confidence
FrontierMath v2 (Tier 4)22.0%measured
FrontierMath v2 (Tiers 1-3)68.4%measured
FrontierSWE v219.6%measured
Furniture Assembly31.7%measured
MaxxBench53.4%measured
MazeBench4.0%measured
MirrorCode38.9%measured
Mystery Game Puzzles47.0%measured
NYT Connections97.4%measured
OfficeQA Pro67.6%estimated ± 6.8 pp, medium confidence
OSWorld-Verified81.3%estimated ± 2.3 pp, medium confidence
ParseBench-171.2%measured
Terminal-Bench 4.118.2%measured
Terminal-Bench Hard57.5%estimated ± 2.8 pp, medium confidence
Vals SWE-bench80.0%measured
Terminal-Bench 2.1 (Vals AI)81.3%measured
VulcanBench v378.2%estimated ± 3.9 pp, medium confidence
CWE-Bench-AA52.5%measured
AA Cyber Index25.1%measured
DeepsecBench-AA14.5%measured
Harvey LAB-AA v1.17.4%estimated ± 4.0 pp, low confidence
HealthBench Professional (length-adjusted, Artificial Analysis)55.2%estimated ± 4.3 pp, low confidence
HealthBench Professional (raw, Artificial Analysis)64.4%estimated ± 0.8 pp, low confidence
AA ITBench52.5%measured
Bug Hunt Bench18.0%measured
Chartography (tools)86.7%estimated ± 4.1 pp, low confidence
CursorBench 4.042.0%estimated ± 2.0 pp, high confidence
CWE-bench v157.4%estimated ± 5.1 pp, medium confidence
DeepSWE73.8%measured
DRACO86.9%estimated ± 0.7 pp, low confidence
FrontierCode 2.024.3%estimated ± 4.8 pp, medium confidence
FrontierPhysics0.0%measured
HLE w/o tools51.4%estimated ± 4.9 pp, high confidence
OSWorld 2.059.0%measured
OSWorld 2.176.3%estimated ± 3.6 pp, high confidence
OSWorld 2.1 Offline50.3%estimated ± 1.9 pp, medium confidence
PostTrainBench v1.124.1%estimated ± 3.7 pp, high confidence
ProgramBench88.0%estimated ± 6.0 pp, low confidence
SpireClimb16.8%estimated ± 7.2 pp, medium confidence
StealthMark47.9%measured
SWE-bench Multimodal56.9%estimated ± 5.6 pp, low confidence
Terminal-Bench 4.019.1%measured
Terminal-Bench-Science 0.10.0%estimated ± 10.7 pp, low confidence
Toolathlon-Verified79.3%estimated ± 1.0 pp, medium confidence
VulcanBench Frontier v490.1%estimated ± 2.3 pp, medium confidence
AA AutomationBench51.2%estimated ± 5.5 pp, medium confidence
GDP.pdf18.9%estimated ± 3.1 pp, high confidence
AA Terminal-Bench 4.017.8%estimated ± 6.2 pp, medium confidence
ARC-AGI-310.4%measured
ARC-AGI-3 Provider Adapter35.0%measured
cursorBench4144.8%estimated ± 3.7 pp, high confidence
CWE-bench47.2%estimated ± 4.3 pp, medium confidence
MKA35.3%measured
SkillsBench official leaderboard (BenchFlow)58.6%estimated ± 6.1 pp, medium confidence
BinaryAudit40.5%estimated ± 8.9 pp, low confidence
OTelBench22.5%estimated ± 6.6 pp, low confidence
Vibe Code Bench76.2%estimated ± 10.0 pp, low confidence
FACTS Multimodal57.6%measured
Hermes Bench62.2%measured
Hermes Index38.1%measured
ParseBench-276.7%measured
CharXiv91.2%estimated ± 2.4 pp, high confidence
CharXiv w/o tools86.2%measured
ExploitBench49.7%estimated ± 1.9 pp, medium confidence
ExploitGym19.5%estimated ± 2.7 pp, high confidence
AI2D_TEST94.2%estimated ± 2.0 pp, low confidence
C-Eval94.8%estimated ± 1.0 pp, low confidence
CountBench100.0%estimated ± 2.9 pp, low confidence
DynaMath85.5%estimated ± 3.0 pp, low confidence
ERQA73.7%estimated ± 2.9 pp, medium confidence
HMMT Feb 202691.8%estimated ± 1.5 pp, medium confidence
HMMT Nov 202595.4%estimated ± 1.5 pp, low confidence
MathVision94.8%estimated ± 3.1 pp, medium confidence
MCP Atlas80.8%estimated ± 5.8 pp, low confidence
MMAnswerBench91.3%estimated ± 2.4 pp, low confidence
MMLU-Redux97.3%estimated ± 0.8 pp, low confidence
MMMU-Pro85.0%estimated ± 1.5 pp, medium confidence
NL2Repo45.5%estimated ± 4.7 pp, high confidence
PostTrainBench v124.4%estimated ± 3.6 pp, high confidence
QwenClawBench93.2%estimated ± 6.0 pp, low confidence
RealWorldQA87.9%estimated ± 1.8 pp, medium confidence
ScreenSpot Pro90.2%estimated ± 10.1 pp, low confidence
SuperGPQA99.9%estimated ± 9.3 pp, low confidence
Terminal-Bench 2.079.8%estimated ± 5.8 pp, low confidence
Toolathlon56.7%estimated ± 1.7 pp, medium confidence
V*100.0%estimated ± 7.5 pp, low confidence
Video-MME (w/o subtitle)90.9%estimated ± 1.7 pp, low confidence
Video-MME (with subtitle)91.6%estimated ± 3.7 pp, low confidence
AA LiveCodeBench91.6%estimated ± 2.6 pp, medium confidence
HealthBench Hard32.2%estimated ± 8.6 pp, low confidence
MedXpertQA (MM)77.8%estimated ± 6.5 pp, low confidence
React Native Evals83.3%estimated ± 3.2 pp, medium confidence
ResearchClawBench20.3%estimated ± 1.9 pp, high confidence
PostTrainBench33.1%estimated ± 2.4 pp, high confidence
DAYJOB2.6%measured
DAYJOB Healthcare1.2%measured
Finance Agent v261.4%measured
BioMysteryBench (human-solvable)88.8%measured
FrontierCode 1.1 Extended62.3%estimated ± 1.6 pp, medium confidence
HealthBench (raw)55.6%estimated ± 6.0 pp, medium confidence
HealthBench (length-adjusted)53.9%estimated ± 0.9 pp, medium confidence
HealthBench Professional59.4%estimated ± 2.1 pp, medium confidence
HealthBench Professional (raw)62.2%estimated ± 2.3 pp, medium confidence
HLE-Verified54.9%measured
LABBench286.2%measured
ScarfBench66.0%measured
BlenderWorld (text-only)77.7%estimated ± 2.8 pp, medium confidence
Cookie-Bench87.3%estimated ± 1.3 pp, medium confidence
STEMBench2.6%estimated ± 0.2 pp, medium confidence
SimpleQA36.3%estimated ± 7.8 pp, low confidence
Agents' Last Exam39.0%estimated ± 9.7 pp, medium confidence
BabyVision w/ Python90.6%estimated ± 3.7 pp, low confidence
OpenHarmony Bench v1.058.9%estimated ± 2.4 pp, high confidence
OpenHarmony Bench v1.252.8%estimated ± 11.0 pp, low confidence
SEC-Bench Pro85.2%estimated ± 5.6 pp, medium confidence
BabyVision70.6%estimated ± 7.4 pp, low confidence
PerceptionBench65.2%estimated ± 3.8 pp, medium confidence
ZeroBench25.9%estimated ± 7.1 pp, low confidence
MRCRv285.9%estimated ± 11.3 pp, low confidence
LVBench87.1%measured
MMLongBench-Doc61.5%estimated ± 2.6 pp, low confidence
SWE-Marathon41.3%estimated ± 2.1 pp, low confidence
BioBasicsBench38.7%estimated ± 2.8 pp, medium confidence
LifeVal77.8%estimated ± 2.3 pp, low confidence
MMMU-Pro w/ Python87.8%estimated ± 2.0 pp, medium confidence
ExploitBench Internal Port5.8%estimated ± 1.4 pp, medium confidence
MathVision w/ Python97.0%estimated ± 1.3 pp, low confidence
SWE-Atlas Codebase QnA61.4%estimated ± 4.8 pp, high confidence