Gemini 3.8 Flash
- Best result
- 98.89onOTIS Mock AIME 2024-2025unverified
- Standing
- #28 of 170 on OTIS Mock AIME 2024-2025 — top quarter of the field
- Leads
- 1 of 23 benchmarksARC-AGI
- Evidence
- 0 of 23 scores independently reproduced
- third-party
- unverified
Each bar is drawn on its own benchmark's scale — the ceiling that benchmark states where it has one (of 100), otherwise the best score recorded on it (rel. 2103). A benchmark that states neither draws no bar.
Leads on: ARC-AGI
| Benchmark | Score | Evidence status | Released |
|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.89accuracy (%) | unverifiedT2 | 2026-09-02 |
| ARC-AGI | 98.5accuracy (%) | third-partyT1 | 2026-09-02https://arcprize.org/leaderboard |
| GPQA diamond | 97.77accuracy (%) | unverifiedT2 | 2026-09-02 |
| DTBench | 92.87accuracy (%) | third-partyT1 | 2026-09-02https://conceptualreasoning.ai/dtbench |
| ARC-AGI-2 | 89.17accuracy (%) | unverifiedT2 | 2026-09-02 |
| GeoBench | 86accuracy (%) | third-partyT1 | 2026-09-02https://geobench.org/ |
| WeirdML | 84.81accuracy (%) | third-partyT1 | 2026-09-02https://htihle.github.io/weirdml.html |
| SimpleBench | 78.88accuracy (%) | third-partyT1 | 2026-09-02https://simple-bench.com/ |
| Surface Evolver Bench | 76.88accuracy (%) | third-partyT1 | 2026-09-02https://yhenon.github.io/surface-evolver-llm-eval/ |
| FrontierMath-Tiers-1-3-v2-Private | 72.02accuracy (%) | unverifiedT2 | 2026-09-02 |
| SimpleQA Verified | 69.7accuracy (%) | unverifiedT2 | 2026-09-02 |
| APEX-Agents | 64.3accuracy (%) | unverifiedT2 | 2026-09-02 |
| LMCA | 62.24accuracy (%) | third-partyT1 | 2026-09-02https://conceptualreasoning.ai/lmca |
| Chess Puzzles | 55.81accuracy (%) | unverifiedT2 | 2026-09-02 |
| ProofBench | 48accuracy (%) | third-partyT1 | 2026-09-02https://www.vals.ai/benchmarks/proof_bench |
| Mystery Game Puzzles | 41.62accuracy (%) | unverifiedT2 | 2026-09-02 |
| FrontierCode | 41.19accuracy (%) | third-partyT1 | 2026-09-02https://cognition.com/frontiercode |
| MirrorCode | 38.89accuracy (%) | unverifiedT2 | 2026-09-02 |
| FrontierMath-Tier-4-v2-Private | 21.95accuracy (%) | unverifiedT2 | 2026-09-02 |
| FrontierSWE | 19.63accuracy (%) | third-partyT1 | 2026-09-02https://www.frontierswe.com/ |
| WeirdML v3 | 16.19accuracy (%) | third-partyT1 | 2026-09-02https://htihle.github.io/weirdml.html |
| Remote Labor Index | 5.83accuracy (%) | unverifiedT2 | 2026-09-02 |
| Furniture Assembly | 2.38accuracy (%) | unverifiedT2 | 2026-09-02 |
Claims drawn from cited facts, not live model generation.
This model was developed by Google DeepMind and released in 2026.
2 cited facts
It is scored on 23 tracked benchmarks, and it currently holds 1 top score. Its record is unverified: outside evaluation has neither confirmed nor disputed these numbers, so treat them as awaiting independent validation.
3 cited facts
Across its benchmarks it trails the leading score by an average of 27.7 points under the disclosed evaluation harness, a wide gap that reads as well behind the front of the pack rather than at the frontier. It nevertheless holds the top score on 1 benchmark, so its standing is not entirely without category leadership, though that lone leader result sits alongside the wide average shortfall.
3 cited facts
1 facts cross-checked across data sources: 1 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks
Gemini 3.8 Flash is an AI model developed by Google DeepMind, released 2026-09-02. Its benchmark record below tags every score as independently reproduced, third-party, self-reported, or unverified.
Gemini 3.8 Flash has recorded scores on 23 benchmarks, each shown with its evidence status.
Gemini 3.8 Flash has recorded scores on 23 benchmarks — ARC-AGI-2, ProofBench, GPQA diamond, OTIS Mock AIME 2024-2025, DTBench, WeirdML, and 17 more. The full table above shows each score with its evidence status.
0 of 23 tracked claims (0%) are independently reproduced; the rest are third-party, self-reported, unverified, or contradicted by an independent measurement.
Gemini 3.8 Flash has 23 tracked claims: 12 unverified.
Plain text
tensor.news. "Gemini 3.8 Flash — benchmark record". Benchmark record with evidence status for every score Updated 2026-11-19. https://tensor.news/models/gemini-3-8-flash
BibTeX
@misc{tensornews_model_gemini-3-8-flash-benchmark-record,
title = {{Gemini 3.8 Flash — benchmark record}},
author = {{tensor.news}},
year = {2026},
month = {11},
howpublished = {\url{https://tensor.news/models/gemini-3-8-flash}},
note = {Benchmark record with evidence status for every score},
urldate = {2026-11-19},
}