Omni-Persona - Text-to-Any: leaderboard

Text-to-Any persona matching on Omni-Persona (2026): a text query is matched to image, text and audio context to answer personalized questions, scored by calibrated accuracy, which combines judged answers on answerable items with correct abstention on items whose persona is absent from the context. Percent; higher is better.

Metric: Calibrated accuracy (%): item-weighted accuracy over answerable items (answer judged correct by a GPT-5.4-mini judge) and absent-persona items (correct abstention by keyword matching), on the text-to-any persona matching group (text query, image, text and audio context), zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1MiniCPM-o-4.5Thinking51.4#1252
2Phi-4 Multimodal Instruct50#4154
3Qwen2.5-Omni-3B45.9#3519
4Qwen3 Omni 30B A3B Instruct44.6#930
5Gemini 3.5 Flash Lite43.2#299
6Interactive-Omni-8B42.6
7gemma-4-E4B-it41.2#1539
8Qwen2.5-Omni-7B40.5#1884
9Gemini 3.6 FlashMedium39.2#81
10Gemma 4 E2B (IT)36.5#2451

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=omni-persona-text-to-any · How It Works · Data refreshed daily, snapshot 2026-11-21.