Model profile
Llama 3.1 8B Base
Evidence summary
Safety. Llama 3.1 8B Base has an estimated Safety rank of #331; its 90% source-sensitivity interval is #286–#333. Its behavior-only rank is #325; company governance moves the combined estimate to #331. Published Safety evidence spans 3 eval lineages and 4 of 7 components. Its strongest relative result is AbstentionBench (stale_f1, #16 of 20); its weakest is AbstentionBench (answer_unknown_f1, #20 of 20).
Freedom. Llama 3.1 8B Base has an estimated Freedom rank of #213; its 90% source-sensitivity interval is #46–#519. Published Freedom evidence spans 1 eval lineages and 1 of 1 components. Its strongest relative result is MedSafetyBench (medical_safety_score, #5 of 30); its weakest is MedSafetyBench (medical_safety_score, #5 of 30).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Official model page ↗Family-level model document · meta · first party
- Release source ↗exact first party family release
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 |
|---|---|---|---|---|
| AbstentionBenchanswer_unknown_f1 | #20 / 20 | ↑0.6322 | Source ↗official | |
| AbstentionBenchfalse_premise_f1 | #19 / 20 | ↑0.474 | Source ↗official | |
| AbstentionBenchstale_f1 | #16 / 20 | ↑0.5877 | Source ↗official | |
| AbstentionBenchsubjective_f1 | #18 / 20 | ↑0.5653 | Source ↗official | |
| AbstentionBenchunderspecified_context_f1 | #16 / 20 | ↑0.4863 | Source ↗official | |
| AbstentionBenchunderspecified_intent_f1 | #20 / 20 | ↑0.5494 | Source ↗official | |
| IndoBias-Pairs — parity-aware culturally grounded biasparity_score | #24 / 26 | ↑81.67 | Source ↗official | |
| MedSafetyBenchmedical_safety_score | #26 / 30 | ↑54.35 | 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 |
|---|---|---|---|---|
| MedSafetyBenchmedical_safety_score | #5 / 30 | ↓54.35 | Source ↗official |