Benchmark Almanac

Qwen3-8B-Base

Qwen3-8B-Base ranks first on EusExams.

Qwen Released April 28th, 2025 Family Qwen3 The lab's model card Compare with others
14
Benchmarks with a score
1
Ranked first, out of five or more models
3
In the top three
Qwen Apr 2025

Qwen3-8B-Base

Open weights

No Epoch Capabilities Index score yet
  • Alyah 40.8% #11 /16
  • QIMMA 56.3% #15 /44
  • EusExams 51.3% #1 /24

Where has it been number one?

Qwen3-8B-Base led 1 benchmark at some point, for a median of 572 days. It still tops 1 that nobody has tested since.

Source: Official leaderboards

How does it do against its rivals?

Qwen3-8B-Base beats Gemma 3 1B PT on 8 of the 8 benchmarks they share. Gemma 3 12B PT does best against it, ahead on 5 of 8.

  1. Gemma 3 1B PT 8 ahead, 0 behind Compare
  2. Gemma 3 4B PT 8 ahead, 2 behind Compare
  3. Gemma 3 4B 6 ahead, 3 behind Compare
  4. Gemma 3 12B PT 3 ahead, 5 behind Compare

Sources: Official leaderboards, Epoch AI (CC BY 4.0) and labs’ model cards

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
EusExams 51.3% #1 of 24 Official leaderboard
EusTrivia 48.3% #2 of 24 Official leaderboard
NOVA-63 56.7% #5 of 52 Official leaderboard
EusReading 46.9% #3 of 24 Official leaderboard
HeQ 78.9% #18 of 88 Official leaderboard
3LM 80.2% #8 of 25 Official leaderboard
Open Hebrew LLM Leaderboard 59.4 #29 of 88 Official leaderboard
QIMMA 56.3% #15 of 44 Official leaderboard
MultiBLiMP 85.8% #9 of 25 Official leaderboard
Uncheatable Eval 7.10 #14 of 33 Official leaderboard
NorEval 52.4 #11 of 19 Official leaderboard
Open Portuguese LLM Leaderboard 62.8 #106 of 175 Official leaderboard
Alyah 40.8% #11 of 16 Official leaderboard
BLUEX 36% #134 of 177 Official leaderboard

Who measured it?

Independent runs cover all 14 of its benchmarks.

  • An independent run only 14
Official leaderboards scored it on 14.

Sources: Official leaderboards, Epoch AI (CC BY 4.0) and labs’ model cards