Model profile
Gemini 2.5 Pro
Googledeveloper
2025-03-25release date
#104 / 267overall rank
33eval lineages
Evidence summary
Gemini 2.5 Pro has an estimated overall rank of #104; its 90% source-sensitivity interval is #55–#173. Its behavior-only rank is #111; company governance moves the combined estimate to #104. Published evidence spans 33 evals and 7 of 7 behavior components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 102); its weakest is PropensityBench (score, #14 of 14).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemini-2-5-pro
- OpenRouter ↗google/gemini-2.5-pro
- 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 | Better | Source |
|---|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #225 / 311 | 0.8738 | ↓ lower | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #30 / 80 | 0.736 | ↑ higher | Source ↗official | |
| Alignment Leaderboardcorrigibility | #9 / 24 | 4.34 | ↑ higher | Source ↗official | |
| Alignment Leaderboardhonesty | #16 / 24 | 3.502 | ↑ higher | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #13 / 24 | 3.46 | ↑ higher | Source ↗official | |
| Alignment Leaderboardrobustness | #16 / 24 | 3.28 | ↑ higher | Source ↗official | |
| Alignment Leaderboardsafety | #13 / 24 | 3.827 | ↑ higher | Source ↗official | |
| Alignment Leaderboardscheming | #11 / 24 | 3.789 | ↑ higher | Source ↗official | |
| Anthropic Agentic Misalignment — blackmailmisaligned_action_rate_pct | #14 / 16 | 95 | ↓ lower | Source ↗official | |
| Anthropic Agentic Misalignment — corporate espionagemisaligned_action_rate_pct | #14 / 16 | 100 | ↓ lower | Source ↗official | |
| Anthropic Agentic Misalignment — lethal actionmisaligned_action_rate_pct | #8 / 10 | 90 | ↓ lower | Source ↗official | |
| BrokenMathsycophancy | #3 / 9 | 37.5 | ↓ lower | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #62 / 105 | 0.2 | ↑ higher | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #38 / 43 | 91.4 | ↓ lower | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #44 / 48 | 94.1 | ↓ lower | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #37 / 49 | 72 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmachiavelli | #37 / 45 | 95.3 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmask | #37 / 51 | 44.3 | ↓ lower | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #5 / 32 | 37.3 | ↓ lower | Source ↗official | |
| CAIS Risk Indextextquests_harm | #15 / 48 | 15.9 | ↓ lower | Source ↗official | |
| Confabulationsconfabulation_rate | #7 / 52 | 4.95 | ↓ lower | Source ↗official | |
| Emergent Collusionhigh_illegality_game_rate | #7 / 13 | 0.39 | ↓ lower | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #38 / 260 | 38.85 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #220 / 260 | 79.17 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #209 / 260 | 45.37 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #139 / 258 | 96.02 | ↑ higher | Source ↗official | |
| FlagEval Safety and Valuesa1_qualified_rate | #3 / 18 | 84.87 | ↑ higher | Source ↗official | |
| FlagEval Safety and Valuesa2_qualified_rate | #4 / 18 | 81.19 | ↑ higher | Source ↗official | |
| FlagEval Safety and Valuesa3_qualified_rate | #11 / 18 | 86.5 | ↑ higher | Source ↗official | |
| FlagEval Safety and Valuesa4_qualified_rate | #4 / 18 | 90.01 | ↑ higher | Source ↗official | |
| FlagEval Safety and Valuesa5_qualified_rate | #2 / 18 | 74.97 | ↑ higher | Source ↗official | |
| FORTRESSaverage_risk_score | #47 / 49 | 61.68 | ↓ lower | Source ↗official | |
| FORTRESSover_refusal_score | #8 / 46 | 1.58 | ↓ lower | Source ↗official | |
| HELM Safetyanthropic_red_team | #18 / 80 | 0.995 | ↑ higher | Source ↗official | |
| HELM Safetybbq | #18 / 80 | 0.964 | ↑ higher | Source ↗official | |
| HELM Safetyharmbench | #53 / 80 | 0.654 | ↑ higher | Source ↗official | |
| HELM Safetysimple_safety_tests | #58 / 80 | 0.97 | ↑ higher | Source ↗official | |
| HELM Safetyxstest | #3 / 80 | 0.987 | ↑ higher | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #4 / 54 | 29.16 | ↑ higher | Source ↗official | |
| LLM Ethics Benchmarkscore | #3 / 5 | 86.1 | ↑ higher | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #39 / 50 | 95.3 | ↓ lower | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #27 / 31 | 9 | ↓ lower | Source ↗official | |
| Manager Coercion Benchfabrication_rate | #12 / 13 | 67 | ↓ lower | Source ↗official | |
| MASKlying_probability_pct | #35 / 53 | 45.63 | ↓ lower | Source ↗official | |
| PHAREbias_resistance_diagnostic | #52 / 66 | 0.3652 | ↑ higher | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #37 / 70 | 0.7482 | ↑ higher | Source ↗official | |
| PHAREharm_resistance_diagnostic | #42 / 70 | 0.9218 | ↑ higher | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #53 / 67 | 0.3867 | ↑ higher | Source ↗official | |
| PropensityBenchscore | #14 / 14 | 79 | ↓ lower | Source ↗official | |
| Shelleducation_jsr | #3 / 14 | 0.418 | ↓ lower | Source ↗official | |
| Shellfinance_jsr | #6 / 14 | 0.4 | ↓ lower | Source ↗official | |
| Shellmanagement_jsr | #5 / 14 | 0.502 | ↓ lower | Source ↗official | |
| Social Welfare Function Benchmarkfairness | #17 / 19 | 0.444 | ↑ higher | Source ↗official | |
| SOSBenchbiology_pvr | #10 / 23 | 0.354 | ↓ lower | Source ↗official | |
| SOSBenchchemistry_pvr | #11 / 23 | 0.342 | ↓ lower | Source ↗official | |
| SOSBenchmedicine_pvr | #10 / 23 | 0.492 | ↓ lower | Source ↗official | |
| SOSBenchpharmacology_pvr | #7 / 23 | 0.634 | ↓ lower | Source ↗official | |
| SOSBenchphysics_pvr | #12 / 23 | 0.466 | ↓ lower | Source ↗official | |
| SOSBenchpsychology_pvr | #7 / 23 | 0.294 | ↓ lower | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 102 | 7 | ↑ higher | Source ↗official | |
| SpeciEvalland_animal_4ns | #63 / 102 | 4.65 | ↓ lower | Source ↗official | |
| SpeciEvalsea_animal_4ns | #47 / 102 | 4.72 | ↓ lower | Source ↗official | |
| SpeciEvalspeciesism | #2 / 102 | 1.05 | ↓ lower | Source ↗official | |
| TrustLLM contemporary collapsed applicationtrustllm | #2 / 8 | 0.63 | ↑ higher | 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.