Model profile
Gemini 2.0 Flash
Evidence summary
Gemini 2.0 Flash has an estimated overall rank of #187; its 90% source-sensitivity interval is #75–#246. Its behavior-only rank is #199; company governance moves the combined estimate to #187. Published evidence spans 13 evals and 6 of 7 behavior components. Its strongest relative result is PandaBench JBB direct-request panel (safety_rate, #1 of 46); its weakest is PropensityBench (score, #13 of 14).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemini-2-0-flash
- OpenRouter ↗google/gemini-2.0-flash-001
- Official model documentation ↗Family-level model document · google · first party
- Release source ↗direct research preview identity date
Published eval results
Rank is within that sub-eval. Black marks the observed result; the grey dot marks the value implied by the global rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #210 / 330 | ↓0.8624 | Source ↗official | |
| AgentDojotargeted_attack_success_rate | #9 / 15 | ↓0.1892 | Source ↗official | |
| AgentDojoutility_under_attack | #7 / 15 | ↑0.3982 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #74 / 106 | ↑0.15 | Source ↗official | |
| Confabulationsconfabulation_rate | #23 / 52 | ↓19.55 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #150 / 241 | ↑13.7 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #35 / 241 | ↑93.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #131 / 241 | ↑63.89 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #169 / 239 | ↑93.64 | Source ↗official | |
| Google Gemini 2.5 Flash Model Cardtext_safety_delta | #1 / 2 | ↓0 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #27 / 54 | ↑27.08 | Source ↗official | |
| Humanity's Last Exam RMS calibration error (Scale Labs)calibrationError | #29 / 39 | ↓82 | Source ↗official | |
| MASKlying_probability_pct | #39 / 53 | ↓49.93 | Source ↗official | |
| PandaBench JBB direct-request panelsafety_rate | #1 / 46 | ↑1 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #18 / 66 | ↑0.5351 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #42 / 70 | ↑0.7159 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #31 / 70 | ↑0.943 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #56 / 67 | ↑0.3735 | Source ↗official | |
| PropensityBenchscore | #13 / 14 | ↓77.8 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #88 / 113 | ↑6.5 | Source ↗official | |
| SpeciEvalland_animal_4ns | #27 / 113 | ↓4.29 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #66 / 113 | ↓4.79 | Source ↗official | |
| SpeciEvalspeciesism | #80 / 113 | ↓2.33 | Source ↗official |
Values evaluations
Descriptive values and political-framing results are separate from safety/ethics ranks. Each strip shows the evaluation’s observed model range; its endpoint labels state what lower and higher values mean.
ValueCompass
| Dimension | Value | Distribution |
|---|---|---|
| Universalism | 67.1 | |
| Self-direction | 48.9 | |
| Care / Harm | 45.1 | |
| Fairness / Cheating | 42.5 | |
| Ethical | 88.9 |