Rankings / Google DeepMind
Gemini 2.0 Flash
released 2025-02-02
BenchAtlas Index
as of 2026-07-14
base configuration
rank #382
7 families · 4 categories · medium
Benchmark evidence
12 results
coding
| LiveCodeBench v6 implementation=artificial-analysis | 21.0 % | independent | Artificial Analysis | |
| SciCode | 34.0 % | independent | Artificial Analysis |
external indices
| Intelligence Index v4.1 | 10.7 points | independent | Artificial Analysis |
factuality
| MASK | 49.1 % 47.1–51.1 | independent | Scale Labs |
knowledge science
| GPQA Diamond implementation=artificial-analysis | 63.6 % | independent | Artificial Analysis | |
| Humanity's Last Exam implementation=artificial-analysis | 4.7 % | independent | Artificial Analysis | |
| MMLU-Pro implementation=artificial-analysis | 78.2 % | independent | Artificial Analysis |
long context instruction
| MultiChallenge | 36.4 % 35.6–37.1 | independent | Scale Labs |
reasoning math
| AIME implementation=artificial-analysis | 30.0 % | independent | Artificial Analysis | |
| ARC-AGI-2 split=semi_private · model_type=Base LLM | 1.3 % | independent | ARC Prize Leaderboard | |
| ARC-AGI-2 split=semi_private · model_type=Base LLM | 0.0 usd_per_task | independent | ARC Prize Leaderboard | |
| MATH-500 implementation=artificial-analysis | 91.1 % | independent | Artificial Analysis |
Agent + model results
systems, not bare-model scores
| agent + model mini-SWE-agent + Gemini 2.0 Flash | SWE-bench bash-only | 13.5 % | community | SWE-bench Leaderboard |
| agent + model mini-SWE-agent + Gemini 2.0 Flash | SWE-bench Verified | 13.5 % | community | SWE-bench Leaderboard |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.