ERNIE-4.5-21B-A3B (Thinking): benchmark results
Baidu ERNIE 4.5 21B A3B evaluated with thinking enabled. Provider: Baidu. Released 2025-06-28. Access: Open.
Aggregate Rating 1449 ± 40, rank #1255 of 2168 rated models, from 15 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AI Chess Leaderboard (Reasoning) | 634 | Elo | 39.4 |
| OpenCompass Research - AIME 2025 | 76.2 | Score (%) | 34 |
| SnakeBench | 18.4 | TrueSkill Rating | 33.6 |
| Wolfram LLM Benchmarking Project | 30.6 | Correct Functionality (%) | 31.6 |
| OpenCompass Research - LiveCodeBench v6 | 47 | Score (%) | 19.1 |
| OpenCompass | 57 | Average Score | 18.1 |
| OpenCompass Research - HLE | 6.5 | Score (%) | 17 |
| ReasonScape M12X | 106.43 | ReasonScore (hard) | 16.5 |
| Tinybird AI SQL Benchmark - First-Attempt Success Rate | 76 | Questions answered with a valid query on the first attempt ( | 14.3 |
| OpenCompass Research - IFEval | 81.2 | Score (%) | 10.6 |
| Tinybird AI SQL Benchmark - Exactness | 30.68 | Result exactness vs human reference queries (0-100) | 10.3 |
| LiveSecBench | 23.66 | Overall Score (%) | 9.5 |
Interactive version: theaggregate.ai/model?slug=ernie-4-5-21b-a3b-thinking · How It Works · Data refreshed daily, snapshot 2026-10-20.