Google / MODEL EVIDENCE
96 benchmarks · 141 observations · 109 sources · released 2026-09-02
96 benchmarks · large dot: Gemini 3.8 Flash; small dots: closest models where reported
Gemini 3.8 Flash (High) · Intelligence Index v5.0 ↗
40.93 index · Gemini 3.8 Flash (High) · Intelligence Index v4.3.2 · Artificial Analysis ↗
58.7 · Gemini 3.8 Flash · AA Intelligence Index v4.1.1 · historical · GPT-6 Astra launch report ↗
Gemini 3.8 Flash (High) · AA-Briefcase v1.1 · overall Elo ↗
Gemini 3.8 Flash (High) · AutomationBench-AA · guardrail-gated objective score ↗
Gemini 3.8 Flash (High) · GDP.pdf · all-pass ↗
23.4% · Gemini 3.8 Flash (medium) · GDP.pdf · GDP.pdf score · % · Epoch AI · GDP.pdf ↗
23.2% · Gemini 3.8 Flash (high) · GDP.pdf · GDP.pdf score · % · Epoch AI · GDP.pdf ↗
Gemini 3.8 Flash (High) · Terminal-Bench 4.0 · AA harness ↗
19.7 % · Gemini 3.8 Flash (Mini-SWE-agent) · Terminal-Bench 4.0 · Score · Vals AI · Terminal-Bench 4.0 ↗
19.1% · Gemini 3.8 Flash · Terminal-Bench 4.0 · GPT-6 Astra launch report ↗
Gemini 3.8 Flash (High) · Terminal-Bench-Science 0.1 · AA harness ↗
8.57 % · Gemini 3.8 Flash (Mini-SWE-agent) · Terminal-Bench-Science · Score · Vals AI · Terminal-Bench-Science ↗
Gemini 3.8 Flash (High) · FrontierCode 2.0 · AA implementation ↗
Gemini 3.8 Flash (High) · CritPt v2 · hard subset · AA implementation ↗
Gemini 3.8 Flash (High) · HLE-Diamond · AA implementation ↗
34.30% · gemini-3.8-flash · Scale Labs · Score · Scale Labs / SEAL · Scale Labs ↗
Gemini 3.8 Flash (High) · AA-Omniscience v1.1 · error-penalized index ↗
29.55 index · Gemini 3.8 Flash (High) · AA-Omniscience · error-penalized score · Artificial Analysis ↗
Gemini 3.8 Flash (High) · AA-Omniscience v1.1 · accuracy ↗
54.60% · Gemini 3.8 Flash (High) · AA-Omniscience · accuracy · Artificial Analysis ↗
Gemini 3.8 Flash (High) · AA-Omniscience v1.1 · hallucination rate, lower is better ↗
55.18% · Gemini 3.8 Flash (High) · AA-Omniscience · hallucination rate, lower is better · Artificial Analysis ↗
Gemini 3.8 Flash (High) · GDPval-AA v2.1 · Elo ↗
Gemini 3.8 Flash (High) · AA-3D · Elo ↗
Gemini 3.8 Flash (High) · AA-AnalystAgent · strict score ↗
Gemini 3.8 Flash (High) · Harvey LAB-AA v1.1 · all-pass ↗
Gemini 3.8 Flash (High) · FrontierCode 1.1 · AA implementation ↗
56.3% · Gemini 3.8 Flash · FrontierCode 1.1 Extended · GPT-6 Astra launch report ↗
43.6% · Gemini 3.8 Flash · FrontierCode 1.1 Main · GPT-6 Astra launch report ↗
41.19% · Gemini 3.8 Flash (medium) · FrontierCode · Main score · % · Epoch AI · FrontierCode ↗
Gemini 3.8 Flash (High) · AA-LCR v1.1 ↗
Gemini 3.8 Flash (High) · SciCode · AA implementation ↗
56.6% · Gemini 3.8 Flash (high) · SciCode · Score · % · Epoch AI · SciCode ↗
55.09% · Gemini 3.8 Flash (medium) · SciCode · Score · % · Epoch AI · SciCode ↗
54.98% · Gemini 3.8 Flash (low) · SciCode · Score · % · Epoch AI · SciCode ↗
Gemini 3.8 Flash (High) · Humanity’s Last Exam · AA implementation ↗
Gemini 3.8 Flash · GPQA Diamond · Accuracy ↗
95.25% · Gemini 3.8 Flash (High) · GPQA Diamond · AA implementation · Artificial Analysis ↗
95.39% · Gemini 3.8 Flash (high) · GPQA diamond · Best score (across scorers) · % · Epoch AI · GPQA diamond ↗
95.3% · Gemini 3.8 Flash · GPQA Diamond · GPT-6 Astra launch report ↗
Gemini 3.8 Flash (High) · ITBench-AA · root-cause score ↗
Gemini 3.8 Flash · MMMU Pro · Accuracy ↗
85.61% · Gemini 3.8 Flash (High) · MMMU-Pro · AA implementation · Artificial Analysis ↗
Gemini 3.8 Flash (High) · MLCR-AA · overall ↗
Gemini 3.8 Flash (High) · HealthBench · HealthBench Professional · length-adjusted ↗
52.1% · Gemini 3.8 Flash · HealthBench · HealthBench Professional · length-adjusted · GPT-6 Astra launch report ↗
Gemini 3.8 Flash (High) · CritPt v2 · code tools ↗
Gemini 3.8 Flash (High) · τ³-Banking ↗
Gemini 3.8 Flash (High) · Terminal-Bench 2.1 ↗
Gemini 3.8 Flash (High) · CritPt · original ↗
18.29% · Gemini 3.8 Flash (high) · CritPt · Accuracy · % · Epoch AI · CritPt ↗
12.29% · Gemini 3.8 Flash (medium) · CritPt · Accuracy · % · Epoch AI · CritPt ↗
4% · Gemini 3.8 Flash (low) · CritPt · Accuracy · % · Epoch AI · CritPt ↗
Gemini 3.8 Flash · DeepSWE v1.1 ↗
73.83% · Gemini 3.8 Flash (high) · DeepSWE · Pass@1 · % · Epoch AI · DeepSWE ↗
71.02% · Gemini 3.8 Flash (medium) · DeepSWE · Pass@1 · % · Epoch AI · DeepSWE ↗
Gemini 3.8 Flash (high) · FrontierMath-Tiers-1-3-v2-Private · Best score (across scorers) · % ↗
Gemini 3.8 Flash (high) · FrontierMath-Tier-4-v2-Private · Best score (across scorers) · % ↗
Gemini 3.8 Flash (high) · OTIS Mock AIME 2024-2025 · Best score (across scorers) · % ↗
Gemini 3.8 Flash (high) · SimpleQA Verified · Best score (across scorers) · % ↗
Gemini 3.8 Flash (high) · Chess Puzzles · Best score (across scorers) · % ↗
Gemini 3.8 Flash (high) · Mystery Game Puzzles · Best score (across scorers) · % ↗
Gemini 3.8 Flash (high) · ARC-AGI · Score · % ↗
97.5% · Gemini 3.8 Flash (medium) · ARC-AGI · Score · % · Epoch AI · ARC-AGI ↗
90.5% · Gemini 3.8 Flash (low) · ARC-AGI · Score · % · Epoch AI · ARC-AGI ↗
Gemini 3.8 Flash (high) · WeirdML · Accuracy · % ↗
Gemini 3.8 Flash (high) · ARC-AGI-2 · Score · % ↗
82.92% · Gemini 3.8 Flash (medium) · ARC-AGI-2 · Score · % · Epoch AI · ARC-AGI-2 ↗
77.5% · Gemini 3.8 Flash (low) · ARC-AGI-2 · Score · % · Epoch AI · ARC-AGI-2 ↗
Gemini 3.8 Flash (high) · CursorBench · Score · % ↗
37.3% · Gemini 3.8 Flash (medium) · CursorBench · Score · % · Epoch AI · CursorBench ↗
Gemini 3.8 Flash (high) · Surface Evolver Bench · Mean score · % ↗
Gemini 3.8 Flash (medium) · LMCA · Score · % ↗
51.16% · Gemini 3.8 Flash (low) · LMCA · Score · % · Epoch AI · LMCA ↗
Gemini 3.8 Flash (medium) · DTBench · Accuracy · % ↗
94.93% · Gemini 3.8 Flash (low) · DTBench · Accuracy · % · Epoch AI · DTBench ↗
Gemini 3.8 Flash (high) · FrontierMath-Erdos · Best score (across scorers) · % ↗
Gemini 3.8 Flash (high) · Furniture Assembly · Best score (across scorers) · % ↗
Gemini 3.8 Flash (high) · FrontierSWE · Score · % ↗
Gemini 3.8 Flash (high) · WebDev Arena · Elo ↗
Gemini 3.8 Flash (high) · WeirdML v3 · Max performance · % ↗
Gemini 3.8 Flash · General ECI · 26 October 2026 fit ↗
Gemini 3.8 Flash · Vals Index · Accuracy ↗
Gemini 3.8 Flash (high) · Compression · reference-anchored score ↗
0.36 reference score · Gemini 3.8 Flash (high) · LM training · reference-anchored score · Vals RSI Index · LM training ↗
0.15 reference score · Gemini 3.8 Flash (high) · Harness engineering · reference-anchored score · Vals RSI Index · Harness engineering ↗
0.00 reference score · Gemini 3.8 Flash (high) · Post-training · reference-anchored score · Vals RSI Index · Post-training ↗
Gemini 3.8 Flash · Vals Finance Index · Accuracy ↗
Gemini 3.8 Flash · Vals Legal Index · Accuracy ↗
Gemini 3.8 Flash · Vals Science Index · Accuracy ↗
Gemini 3.8 Flash · Legal Research Bench · All-Pass ↗
Gemini 3.8 Flash · HLAB-Vals-1.1 · Task Pass Rate ↗
Gemini 3.8 Flash · LegalBench · Accuracy ↗
Gemini 3.8 Flash · Harvey's Legal Agent Benchmark v1 · Accuracy ↗
Gemini 3.8 Flash · Finance Agent · Accuracy ↗
Gemini 3.8 Flash · Excel Modeling Benchmark · Accuracy ↗
72.2 % · Gemini 3.8 Flash · Excel Modeling Benchmark v1 · Accuracy · Vals AI · Excel Modeling Benchmark v1 ↗
Gemini 3.8 Flash · Tax Agent Bench · Accuracy ↗
Gemini 3.8 Flash · Finance Agent v2 · Accuracy ↗
Gemini 3.8 Flash · MortgageTax · Accuracy ↗
74.45 % · Gemini 3.8 Flash · TaxEval · Accuracy · Vals AI · TaxEval ↗
Gemini 3.8 Flash · MedCode · Accuracy ↗
Gemini 3.8 Flash · MedScribe · Accuracy ↗
Gemini 3.8 Flash · Urgent safety · share of applicable ratings marked fail ↗
69% · Gemini 3.8 Flash · Health advice · share of applicable ratings marked fail · Vals AI · SAFE-Teen ↗
63% · Gemini 3.8 Flash · Relationship boundaries · share of applicable ratings marked fail · Vals AI · SAFE-Teen ↗
Gemini 3.8 Flash · ProofBench · Accuracy ↗
Gemini 3.8 Flash · BioMysteryBench · Accuracy ↗
Gemini 3.8 Flash · MysteryMechanism · Accuracy ↗
Gemini 3.8 Flash · MMLU Pro · Accuracy ↗
Gemini 3.8 Flash · SAGE · Accuracy ↗
Gemini 3.8 Flash · Vibe Code Bench · Vibe Code Bench 1–100 ↗
78.65 % · Gemini 3.8 Flash (OpenHands) · Vibe Code Bench · original task set · Vals AI · Vibe Code Bench ↗
Gemini 3.8 Flash · Code Migration · Accuracy ↗
36.55 % · Gemini 3.8 Flash · Code Migration v1 · Accuracy · Vals AI · Code Migration v1 ↗
Gemini 3.8 Flash · ProgramBench · Accuracy ↗
Gemini 3.8 Flash (OpenHands) · SkillsBench · Score ↗
Gemini 3.8 Flash · IOI · Accuracy ↗
Gemini 3.8 Flash · SWE-bench Multimodal · Accuracy ↗
Gemini 3.8 Flash (Mini-SWE-agent) · SWE-bench Verified · Score ↗
Gemini 3.8 Flash · LiveCodeBench · Accuracy ↗
Gemini 3.8 Flash · CyberBench · Accuracy ↗
Gemini 3.8 Flash · SRE Bench · Accuracy ↗
Gemini 3.8 Flash · CUA-bench · Accuracy ↗
Gemini 3.8 Flash · Time Horizon Index: KSP · Overall ↗
Gemini 3.8 Flash · Public Benefits Bench · Accuracy ↗
Gemini-3.8-Flash (high) · Humanity's Sixth Sense · Score ↗
gemini-3.8-flash · DistressBench · Score ↗
Gemini 3.8 Flash (Mini-SWE-Agent) · SWE Atlas - Refactoring · Score ↗
Gemini 3.8 Flash (Mini-SWE-Agent) · SWE Atlas - Test Writing · Score ↗
Gemini 3.8 Flash (Mini-SWE-Agent) · Scale Labs · Score ↗
Gemini 3.8 Flash · HiL-Bench (Human-in-Loop Benchmark) · Score ↗
Gemini 3.8 Flash (mini-swe-agent) high · SWE-Bench Pro V2 · full ↗
58.80% · Gemini 3.8 Flash (mini-swe-agent) high · SWE-Bench Pro V2 · hard · Scale Labs / SEAL · SWE-Bench Pro V2 ↗
gemini-3.8-flash (high) · Scale Labs · audiomc ↗
60.40% · gemini-3.8-flash (high) · Scale Labs · audiomc-text · Scale Labs / SEAL · Scale Labs ↗
Gemini 3.8 Flash · Scale Labs · Score ↗
Gemini 3.8 Flash · Scale Labs · Score ↗
Gemini 3.8 Flash · Scale Labs · Score ↗
Gemini 3.8 Flash · End-of-year balance ↗
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
60 of 82 shared benchmark comparisons
55 of 82 shared benchmark comparisons
54 of 82 shared benchmark comparisons
Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
Gemini 3.8 Flash · Gemini 3.8 Flash (High) · Gemini 3.8 Flash (Mini-SWE-Agent) · Gemini 3.8 Flash (Mini-SWE-agent) · Gemini 3.8 Flash (OpenHands) · Gemini 3.8 Flash (high) · Gemini 3.8 Flash (low) · Gemini 3.8 Flash (medium) · Gemini 3.8 Flash (mini-swe-agent) high · Gemini-3.8-Flash (high) · gemini-3.8-flash · gemini-3.8-flash (high)