Phi-Line_14B — benchmark results

SicariusSicariiStuff's roleplay and creative-writing fine-tune of Microsoft's Phi-4 14B, tuned for low refusals and character-card adherence. Provider: Microsoft. Released 2025-02-17. Access: Open.

Unified ELO 1550 ± 31, rank #620 of 1776 rated models, from 9 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open LLM Leaderboard - MMLU-Pro49.49Score98.3
Open LLM Leaderboard - MATH Level 538.6Score89.1
Open LLM Leaderboard - GPQA13.76Score88.9
Open LLM Leaderboard - BBH43.79Score83.2
Open LLM Leaderboard - MuSR14.78Score80.2
Open LLM Leaderboard - IFEval64.96Score77.2
UGI - Willingness (W/10)6.8W/10 Score66.6
UGI Leaderboard30.47UGI Score33
UGI - Natural Intelligence16.04NatInt Score24.3

Interactive version: theaggregate.ai/model?slug=phi-line-14b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.