Model profile
Marin 8B Instruct
Marindeveloper
2025-05-14release date
#254 / 267overall rank
5eval lineages
Evidence summary
Marin 8B Instruct has an estimated overall rank of #254; its 90% source-sensitivity interval is #158–#258. Its behavior-only rank is #258; company governance moves the combined estimate to #254. Published evidence spans 5 evals and 3 of 7 behavior components. Its strongest relative result is HELM Safety (harmbench, #39 of 80); its weakest is HELM Safety (bbq, #71 of 80).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗marin-community/marin-8b-instruct
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- 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 |
|---|---|---|---|---|---|
| HELM Safetyanthropic_red_team | #63 / 80 | 0.97 | ↑ higher | Source ↗official | |
| HELM Safetybbq | #71 / 80 | 0.755 | ↑ higher | Source ↗official | |
| HELM Safetyharmbench | #39 / 80 | 0.751 | ↑ higher | Source ↗official | |
| HELM Safetysimple_safety_tests | #66 / 80 | 0.958 | ↑ higher | Source ↗official | |
| HELM Safetyxstest | #67 / 80 | 0.917 | ↑ higher | Source ↗official |