Marin 8B Instruct: benchmark results

Stanford CRFM's Marin 8B Instruct, an SFT-only fine-tune of the open Marin 8B Base on a mix of reasoning, math and instruction datasets including OpenThoughts, Tulu-3 and dolphin-r1. Provider: Stanford CRFM. Released 2025-05-14. Access: Open.

Aggregate Rating 1358 ± 48, rank #7507 of 10556 rated models, from 12 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
HELM Safety - Harm Bench75.06LM Evaluated Safety score53.5
HELM Safety - Safety Scenarios87Mean score25.6
HELM Safety - Anthropic Red Team97.02LM Evaluated Safety score23.3
HELM Safety - Simple Safety Tests95.75LM Evaluated Safety score19.2
SwitzerlandQA53.7Exact match (%)17.6
HELM Safety - Xstest91.67LM Evaluated Safety score15.1
HELM Safety - Bbq75.5BBQ accuracy14
HELM Capabilities - Omni-MATH15.95Acc10.4
HELM Capabilities - IFEval63.15IFEval Strict Acc4.5
HELM Capabilities - MMLU-Pro18.8COT correct1.5
HELM Capabilities - GPQA16.82COT correct0
HELM Capabilities - WildBench47.71WB Score0

Interactive version: theaggregate.ai/model?slug=marin-8b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-17.