# Blind Preference Benchmark v3 Summary (t2i_full_40_20260611_184137)

- task_type: `t2i`
- prompt_file: `/Users/twc/My LAB/202606-MAI-LLM/image/data/prompts/prompts_t2i_40.csv`
- run_status: `official`
- winner_model: `gpt-image-2`
- winner_reason: `ok`
- generated_at: 2026-06-11T19:41:53

## Gates

- PASS `generation_success_gpt`
- PASS `generation_success_mai`
- PASS `judge_valid_rate_100`
- PASS `battle_count_40`
- PASS `no_judge_failure`

## Model Metrics

### gpt
- BPS: 97.5
- wins: 39
- losses: 1
- win_rate: 0.975
- generation_success_rate: 100.0
- judge_valid_rate: 100.0
- avg_generation_time_ms: 35500.93925
- avg_cost_per_image_usd: 0.0

### mai
- BPS: 2.5
- wins: 1
- losses: 39
- win_rate: 0.025
- generation_success_rate: 100.0
- judge_valid_rate: 100.0
- avg_generation_time_ms: 22126.55475
- avg_cost_per_image_usd: 0.0