Model profile
OLMo 2 7B Instruct
Evidence summary
Safety. OLMo 2 7B Instruct has an estimated Safety rank of #319; its 90% source-sensitivity interval is #231–#326. Its behavior-only rank is #327; company governance moves the combined estimate to #319. Published Safety evidence spans 6 eval lineages and 5 of 7 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (harmful_attack_non_success_rate, #67 of 248); its weakest is HELM Safety (bbq, #79 of 80).
Freedom. OLMo 2 7B Instruct has an estimated Freedom rank of #197; its 90% source-sensitivity interval is #53–#572. Published Freedom evidence spans 5 eval lineages and 1 of 1 components. Its strongest relative result is HELM Safety (anthropic_red_team, #3 of 80); its weakest is HELM Safety (xstest, #71 of 80).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Release source ↗direct research preview identity date
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #209 / 248 | ↑10.59 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #93 / 248 | ↑89.67 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #67 / 248 | ↑86.67 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #84 / 246 | ↑97.82 | Source ↗official | |
| HELM Safetyanthropic_red_team | #78 / 80 | ↑0.88 | Source ↗official | |
| HELM Safetybbq | #79 / 80 | ↑0.525 | Source ↗official | |
| HELM Safetyharmbench | #63 / 80 | ↑0.578 | Source ↗official | |
| HELM Safetysimple_safety_tests | #77 / 80 | ↑0.775 | Source ↗official | |
| HELM Safetyxstest | #71 / 80 | ↑0.888 | Source ↗official |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #152 / 248 | ↓89.67 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #180 / 248 | ↓86.67 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #162 / 246 | ↓97.82 | Source ↗official | |
| HELM Safetyanthropic_red_team | #3 / 80 | ↓0.88 | Source ↗official | |
| HELM Safetyharmbench | #18 / 80 | ↓0.578 | Source ↗official | |
| HELM Safetysimple_safety_tests | #4 / 80 | ↓0.775 | Source ↗official | |
| HELM Safetyxstest | #71 / 80 | ↑0.888 | Source ↗official |