Qwen3-8B-Base: benchmark results
Alibaba's open 8B dense pretrained base from the April 2025 Qwen3 family, trained on ~36T tokens across 119 languages. Provider: Alibaba. Released 2025-04-29. Access: Open.
Aggregate Rating 1545 ± 14, rank #626 of 1656 rated models, from 402 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LA Leaderboard - Parafraseja | 68.3 | Accuracy (%) | 100 |
| STEMVerse - Biology - Bloom Level: Evaluate | 83.33 | Few-shot accuracy (%) | 100 |
| STEMVerse - Biology - Immunology | 66.67 | Few-shot accuracy (%) | 100 |
| Ukrainian LLM - Global MMLU Full UK Abstract Algebra | 61 | Score (%) | 100 |
| Ukrainian LLM - Global MMLU Full UK Astronomy | 84.21 | Score (%) | 100 |
| Ukrainian LLM - Global MMLU Full UK College Chemistry | 54 | Score (%) | 100 |
| Ukrainian LLM - Global MMLU Full UK College Computer Science | 70 | Score (%) | 100 |
| Ukrainian LLM - Global MMLU Full UK College Mathematics | 52 | Score (%) | 100 |
| Ukrainian LLM - Global MMLU Full UK College Physics | 62.75 | Score (%) | 100 |
| Ukrainian LLM - Global MMLU Full UK Conceptual Physics | 79.15 | Score (%) | 100 |
| Ukrainian LLM - Global MMLU Full UK Econometrics | 63.16 | Score (%) | 100 |
| Ukrainian LLM - Global MMLU Full UK Electrical Engineering | 75.86 | Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=qwen3-8b-base · How It Works · Data refreshed daily, snapshot 2026-10-20.