← Models

Model profile

Deepseek R1 Distill Llama 70B

DeepSeekdeveloper
2025-01-20release date
#265 / 267overall rank
4eval lineages

Evidence summary

Deepseek R1 Distill Llama 70B has an estimated overall rank of #265; its 90% source-sensitivity interval is #235–#266. Its behavior-only rank is #257; company governance moves the combined estimate to #265. Published evidence spans 4 evals and 3 of 7 behavior components. Its strongest relative result is AA-Omniscience (hallucination_rate, #165 of 311); its weakest is SOSBench (chemistry_pvr, #23 of 23).

Compare this model

Only models sharing at least one published sub-eval are listed.

Official and reference links

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-evalRankValueDistributionBetterSource
AA-Omnisciencehallucination_rate#165 / 3110.8151↓ lowerSource ↗official
AbstentionBenchanswer_unknown_f1#17 / 200.7501↑ higherSource ↗official
AbstentionBenchfalse_premise_f1#16 / 200.6222↑ higherSource ↗official
AbstentionBenchstale_f1#19 / 200.5268↑ higherSource ↗official
AbstentionBenchsubjective_f1#19 / 200.525↑ higherSource ↗official
AbstentionBenchunderspecified_context_f1#13 / 200.5202↑ higherSource ↗official
AbstentionBenchunderspecified_intent_f1#17 / 200.6276↑ higherSource ↗official
PandaBench JBB direct-request panelsafety_rate#28 / 460.97↑ higherSource ↗official
SOSBenchbiology_pvr#20 / 230.838↓ lowerSource ↗official
SOSBenchchemistry_pvr#23 / 230.904↓ lowerSource ↗official
SOSBenchmedicine_pvr#20 / 230.854↓ lowerSource ↗official
SOSBenchpharmacology_pvr#23 / 230.972↓ lowerSource ↗official
SOSBenchphysics_pvr#23 / 230.886↓ lowerSource ↗official
SOSBenchpsychology_pvr#20 / 230.816↓ lowerSource ↗official