Model profile
OLMo 2 32B Instruct
Evidence summary
Safety. OLMo 2 32B Instruct has an estimated Safety rank of #246; its 90% source-sensitivity interval is #81–#288. Its behavior-only rank is #249; company governance moves the combined estimate to #246. Published Safety evidence spans 8 eval lineages and 6 of 7 components. Its strongest relative result is AILuminate General Purpose AI Chat (child_sexual_exploitation, #6 of 32); its weakest is HELM Safety (bbq, #74 of 80).
Freedom. OLMo 2 32B Instruct has an estimated Freedom rank of #452; its 90% source-sensitivity interval is #292–#547. Published Freedom evidence spans 5 eval lineages and 1 of 1 components. Its strongest relative result is HELM Safety (simple_safety_tests, #31 of 80); its weakest is AILuminate General Purpose AI Chat (child_sexual_exploitation, #27 of 32).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗allenai/OLMo-2-0325-32B-Instruct
- OpenRouter ↗allenai/olmo-2-0325-32b-instruct
- 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 |
|---|---|---|---|---|
| AgentDrive Safety Compliancescr | #29 / 48 | ↑76.25 | Source ↗official | |
| AILuminate General Purpose AI Chatchild_sexual_exploitation | #6 / 32 | ↓9.1 | Source ↗official | |
| AILuminate General Purpose AI Chatdefamation | #10 / 32 | ↓10.6 | Source ↗official | |
| AILuminate General Purpose AI Chathate | #8 / 32 | ↓4 | Source ↗official | |
| AILuminate General Purpose AI Chatindiscriminate_weapons_cbrne | #11 / 32 | ↓18.8 | Source ↗official | |
| AILuminate General Purpose AI Chatintellectual_property | #11 / 32 | ↓8.2 | Source ↗official | |
| AILuminate General Purpose AI Chatnon_violent_crimes | #9 / 32 | ↓9 | Source ↗official | |
| AILuminate General Purpose AI Chatprivacy | #10 / 32 | ↓8.1 | Source ↗official | |
| AILuminate General Purpose AI Chatsex_related_crimes | #11 / 32 | ↓9.2 | Source ↗official | |
| AILuminate General Purpose AI Chatsexual_content | #15 / 32 | ↓21 | Source ↗official | |
| AILuminate General Purpose AI Chatspecialized_advice | #12 / 32 | ↓13.9 | Source ↗official | |
| AILuminate General Purpose AI Chatsuicide_self_harm | #10 / 32 | ↓9.7 | Source ↗official | |
| AILuminate General Purpose AI Chatviolent_crimes | #11 / 32 | ↓12.7 | Source ↗official | |
| HELM Safetyanthropic_red_team | #27 / 80 | ↑0.993 | Source ↗official | |
| HELM Safetybbq | #74 / 80 | ↑0.714 | Source ↗official | |
| HELM Safetyharmbench | #31 / 80 | ↑0.841 | Source ↗official | |
| HELM Safetysimple_safety_tests | #48 / 80 | ↑0.98 | Source ↗official | |
| HELM Safetyxstest | #46 / 80 | ↑0.954 | Source ↗official | |
| UAVBench safety-critical decision recognitionethical_safety_critical_accuracy | #21 / 27 | ↑0.615 | 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 |
|---|---|---|---|---|
| AILuminate General Purpose AI Chatchild_sexual_exploitation | #27 / 32 | ↑9.1 | Source ↗official | |
| AILuminate General Purpose AI Chatdefamation | #23 / 32 | ↑10.6 | Source ↗official | |
| AILuminate General Purpose AI Chathate | #25 / 32 | ↑4 | Source ↗official | |
| AILuminate General Purpose AI Chatindiscriminate_weapons_cbrne | #22 / 32 | ↑18.8 | Source ↗official | |
| AILuminate General Purpose AI Chatintellectual_property | #21 / 32 | ↑8.2 | Source ↗official | |
| AILuminate General Purpose AI Chatnon_violent_crimes | #24 / 32 | ↑9 | Source ↗official | |
| AILuminate General Purpose AI Chatprivacy | #23 / 32 | ↑8.1 | Source ↗official | |
| AILuminate General Purpose AI Chatsex_related_crimes | #22 / 32 | ↑9.2 | Source ↗official | |
| AILuminate General Purpose AI Chatsexual_content | #18 / 32 | ↑21 | Source ↗official | |
| AILuminate General Purpose AI Chatspecialized_advice | #21 / 32 | ↑13.9 | Source ↗official | |
| AILuminate General Purpose AI Chatsuicide_self_harm | #22 / 32 | ↑9.7 | Source ↗official | |
| AILuminate General Purpose AI Chatviolent_crimes | #22 / 32 | ↑12.7 | Source ↗official | |
| HELM Safetyanthropic_red_team | #50 / 80 | ↓0.993 | Source ↗official | |
| HELM Safetyharmbench | #49 / 80 | ↓0.841 | Source ↗official | |
| HELM Safetysimple_safety_tests | #31 / 80 | ↓0.98 | Source ↗official | |
| HELM Safetyxstest | #46 / 80 | ↑0.954 | Source ↗official |