Palmyra Fin — benchmark results

Writer's finance-specialist Palmyra model (70B, 32K context), notable as the first model to pass a CFA Level III sample exam (July 2024). Provider: Writer. Released 2024-07-30. Access: API.

Unified ELO 1561 ± 20, rank #588 of 1776 rated models, from 12 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
HELM Safety SimpleSafetyTests100LM Evaluated Safety score (%)86
HELM Safety Anthropic Red Team99.5LM Evaluated Safety score (%)76.7
HELM Safety94.8Mean score (self-reported)71.1
HELM Safety HarmBench84.1LM Evaluated Safety score (%)62.2
HELM Safety BBQ94.2BBQ accuracy (%)55.8
HELM Safety XSTest96.2LM Evaluated Safety score (%)54.7
HELM AIR-Bench66.3Refusal Rate (%)51.2
HELM Capabilities - Omni-MATH29.47Acc44
HELM Capabilities - WildBench78.34WB Score40
HELM Capabilities - GPQA42.15COT correct36
HELM Capabilities - IFEval79.3IFEval Strict Acc34
HELM Capabilities - MMLU-Pro59.1COT correct26

Interactive version: theaggregate.ai/model?slug=palmyra-fin · How the rankings work · Data refreshed daily, snapshot 2026-07-22.