Model profile
Gemini 1.0 Pro
Googledeveloper
2023-12-06release date
#57 / 267overall rank
8eval lineages
Evidence summary
Gemini 1.0 Pro has an estimated overall rank of #57; its 90% source-sensitivity interval is #22–#183. Its behavior-only rank is #65; company governance moves the combined estimate to #57. Published evidence spans 8 evals and 6 of 7 behavior components. Its strongest relative result is OR-Bench (over_refusal_rate, #2 of 25); its weakest is OR-Bench (toxic_acceptance_rate, #22 of 25).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemini-1-0-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 |
|---|---|---|---|---|---|
| AIRBench 2024 Safety Scenariossafety_scenarios | #53 / 80 | 0.582 | ↑ higher | Source ↗official | |
| DecodingTrustmachine_ethics | #1 / 8 | 93.74 | ↑ higher | Source ↗official | |
| DecodingTruststereotype_bias | #2 / 8 | 98.33 | ↑ higher | Source ↗official | |
| DecodingTrusttoxicity | #3 / 8 | 77.53 | ↑ higher | Source ↗official | |
| HarmBenchdr | #11 / 28 | 18 | ↓ lower | Source ↗official | |
| Large-scale Moral Machine experiment on LLMshuman_choice_distance | #12 / 39 | 0.8746 | ↓ lower | Source ↗official | |
| OR-Benchover_refusal_rate | #2 / 25 | 9.7 | ↓ lower | Source ↗official | |
| OR-Benchtoxic_acceptance_rate | #22 / 25 | 21.3 | ↓ lower | Source ↗official | |
| S-Evalbase_en_overall | #19 / 22 | 41.9 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #16 / 33 | 13.36 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #12 / 33 | 20.85 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #16 / 33 | 12.23 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #15 / 33 | 17.93 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #8 / 33 | 27.13 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #16 / 33 | 12.99 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #22 / 33 | 91.09 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #24 / 33 | 91.4 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_malicious_use | #23 / 33 | 87.27 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #16 / 33 | 93.75 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #22 / 33 | 87.37 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #24 / 33 | 82.26 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_human_autonomy_integrity | #11 / 33 | 45.28 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_information_safety_harms | #11 / 33 | 43.33 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_malicious_use | #11 / 33 | 46.6 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_misinformation_harms | #11 / 33 | 44.76 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_representation_toxicity | #15 / 33 | 39.9 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_socioeconomic_harms | #11 / 33 | 44.44 | ↑ higher | Source ↗official | |
| SORRY-Benchavg | #29 / 51 | 0.33 | ↓ lower | Source ↗official |