MiniCPM-o-4.5: benchmark results

Provider: OpenBMB. Released 2026-02-03. Access: Open.

Unified ELO 1606 ± 14, rank #432 of 1605 rated models, from 80 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
DuplexSpeechBench-IFEval (PAS)66.2Persona Adherence Score (0-100; GPT-4o-judged persona-consis100
KMMAU - Topic Summary98Accuracy (%; four-choice question on the topic of the audio;100
MOV-Bench - Hypothetical Reasoning40Accuracy (%; 60 hypothetical-reasoning questions; direct inf100
VideoFDB - Perception3.4gpt-4o judge score (0-5; the 105 held-out test clips of the 100
VideoFDB - Perception (Audio Only)3.44gpt-4o judge score (0-5; the 105 held-out test clips of the 100
VideoFDB - Perception (Audio Only) - Conversational Flow3.76gpt-4o judge score (0-5; the 105 held-out test clips of the 100
VideoFDB - Perception (Audio Only) - Fluency3.45gpt-4o judge score (0-5; the 105 held-out test clips of the 100
VideoFDB - Perception - Conversational Flow3.54gpt-4o judge score (0-5; the 105 held-out test clips of the 100
VideoFDB - Perception - Semantic Grounding3.63gpt-4o judge score (0-5; the 105 held-out test clips of the 100
MathVista82.2Accuracy (%)98.1
HallusionBench65Overall Score97.7
MMStar74.5Accuracy (%)97

Interactive version: theaggregate.ai/model?slug=minicpm-o-4-5 · How It Works · Data refreshed daily, snapshot 2026-09-26.