Qwen2 Instruct (72B) wins more areas, narrowly.
Scores updated · 29 tests both models report · How we compare
Where each one wins
Tests won in each of the two areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
Coding rests on a single test.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Qwen2 Instruct (72B) pulls ahead
- Code for real scientific research problemsSciCode+11points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+7.5points ahead
Where Llama 3 Instruct 8B pulls ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+1.4points ahead
Every test, side by side
All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.
ReasoningEven
- GPQA DiamondQwen2 Instruct (72B) by 7.529.637.1+7.5
- Humanity's Last ExamLlama 3 Instruct 8B by 1.45.13.7+1.4
Other results26 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- MultiPL-EQwen2 Instruct (72B) by 46.622.669.2+46.6
- GSM8KQwen2 Instruct (72B) by 41.249.991.1+41.2
- MATHQwen2 Instruct (72B) by 39.939.179+39.9
- CMMLUQwen2 Instruct (72B) by 39.350.890.1+39.3
- EvalPlusQwen2 Instruct (72B) by 38.740.379+38.7
- C-EvalQwen2 Instruct (72B) by 34.349.583.8+34.3
- MMLU-ProQwen2 Instruct (72B) by 2935.464.4+29
- HumanEvalQwen2 Instruct (72B) by 25.660.486+25.6
- BBHQwen2 Instruct (72B) by 24.757.782.4+24.7
- Theorem QAQwen2 Instruct (72B) by 22.322.144.4+22.3
- WinoGrandeQwen2 Instruct (72B) by 20.16585.1+20.1
- OpenBookQAQwen2 Instruct (72B) by 18.876.695.4+18.8
- bbqQwen2 Instruct (72B) by 18.676.595.1+18.6
- MMLUQwen2 Instruct (72B) by 16.760.176.9+16.7
- HellaSwagQwen2 Instruct (72B) by 16.571.187.6+16.5
- ARC-ChallengeLlama 3 Instruct 8B by 13.982.868.9+13.9
- MBPPQwen2 Instruct (72B) by 12.567.780.2+12.5
- AIR-Bench 2024Llama 3 Instruct 8B by 8.870.962.1+8.8
- TruthfulQALlama 3 Instruct 8B by 8.463.254.8+8.4
- HarmBenchQwen2 Instruct (72B) by 4.172.776.8+4.1
- NarrativeQALlama 3 Instruct 8B by 2.775.472.7+2.7
- AA IntelligenceQwen2 Instruct (72B) by 1.54.86.3+1.5
- XSTestQwen2 Instruct (72B) by 1.395.696.9+1.3
- naturalquestions_closedbookQwen2 Instruct (72B) by 1.237.839+1.2
- simple_safety_teststie99.398.5tie
- anthropic_red_teamtie98.899.1tie
Questions people ask
Which is better, Llama 3 Instruct 8B or Qwen2 Instruct (72B)?
Qwen2 Instruct (72B) wins one of the two areas where both have results: coding. Llama 3 Instruct 8B wins none. They are level on reasoning.
Which is better for coding?
Qwen2 Instruct (72B). It wins the one coding test both models report.
How do you compare the two?
We use the 29 benchmark tests both models have published scores on. The verdict counts the 3 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 26 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.