Google DeepMind·released 2025-12-171 source
Gemini 3 Flash benchmark scores: 21 benchmarks tracked, leading 1. 29% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 1 of 21 benchmarks·best result 95.55 on OTIS Mock AIME 2024-2025 (reproduced)·6 of 21 independently reproduced·$0.5/$3 per M tokens
Head-to-headGemini 3 Flash vs Gemini 2.5 Pro (May 2025)Consensus: LiteLLM · See every model’s pricing →
Leads on: GeoBench
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 95.55accuracy (%) | reproduced· optimizedT1 | 2025-12-17Epoch AI |
| GeoBench | 88accuracy (%) | unverifiedT1 | 2025-12-17GeoBench leaderboard |
| GPQA diamond | 85.86accuracy (%) | reproduced· optimizedT1 | 2025-12-17Epoch AI |
| ARC-AGI | 84.67accuracy (%) | unverified· optimizedT1 | 2025-12-17ARC Prize Leaderboard |
| DTBench | 81.78accuracy (%) | unverifiedT1 | 2025-12-17https://conceptualreasoning.ai/dtbench |
| SWE-Bench verified | 75.41accuracy (%) | reproduced· optimizedT1 | 2025-12-17Epoch AI |
| SimpleQA Verified | 66.8accuracy (%) | unverifiedT2 | 2025-12-17 |
| Terminal Bench | 64.27accuracy (%) | unverifiedT1 | 2025-12-17https://www.tbench.ai/leaderboard/terminal-bench/2.0 |
| WeirdML | 61.6accuracy (%) | unverifiedT1 | 2025-12-17https://htihle.github.io/weirdml.html |
| VPCT | 58.9accuracy (%) | unverifiedT1 | 2025-12-17VPCT leaderboard |
| SimpleBench | 53.32accuracy (%) | unverifiedT1 | 2025-12-17https://lmcouncil.ai/benchmarks |
| FrontierMath-Tiers-1-3-v2-Private | 51.23accuracy (%) | reproducedT1 | 2025-12-17Epoch AI |
| LMCA | 50.74accuracy (%) | unverifiedT1 | 2025-12-17https://conceptualreasoning.ai/lmca |
| DeepResearch Bench | 49.8accuracy (%) | unverifiedT1 | 2025-12-17https://evals.futuresearch.ai/#drb |
| Balrog | 48.1accuracy (%) | unverifiedT1 | 2025-12-17https://balrogai.com/ |
| Chess Puzzles | 35.82accuracy (%) | unverifiedT2 | 2025-12-17 |
| ARC-AGI-2 | 33.61accuracy (%) | unverifiedT2 | 2025-12-17 |
| Mystery Game Puzzles | 18.48accuracy (%) | reproducedT1 | 2025-12-17Epoch AI |
| FrontierMath-Tier-4-v2-Private | 17.07accuracy (%) | reproducedT1 | 2025-12-17Epoch AI |
| ProofBench | 15accuracy (%) | unverifiedT1 | 2025-12-17https://www.vals.ai/benchmarks/proof_bench |
| GSO-Bench | 9.8accuracy (%) | unverifiedT1 | 2025-12-17https://gso-bench.github.io/index.html |
Claims drawn from cited facts, not live model generation.
This model was developed by Google DeepMind and released in 2025.
2 cited facts
This model is scored on 21 tracked benchmarks, and it currently holds the top score on 1 of them. The record’s evidence status is independently reproduced.
3 cited facts
Its mean gap to the best recorded score across its tracked benchmarks is 32.02 points. It currently holds the top score on 1 of them.
2 cited facts
5 facts cross-checked across data sources: 5 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks, litellm_prices
Gemini 3 Flash is an AI model developed by Google DeepMind, released 2025-12-17. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Gemini 3 Flash has recorded scores on 21 benchmarks, each shown with its evidence status.
Gemini 3 Flash has recorded scores on 21 benchmarks — GPQA diamond, OTIS Mock AIME 2024-2025, DTBench, WeirdML, LMCA, SimpleQA Verified, and 15 more. The full table above shows each score with its evidence status.
6 of 21 recorded scores (29%) are independently reproduced rather than self-reported by the lab.
Gemini 3 Flash has 21 tracked claims: 6 independently reproduced, 15 unverified.
Listed API pricing: $0.5 per million input tokens, $3 per million output tokens. See the pricing block for the full breakdown.