OLMo 2 32B wins more areas, narrowly.
Scores updated · 8 tests both models report · How we compare
Where each one wins
Tests won in each of the four 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.
Long documents and following instructions rest on a single test each.
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 OLMo 2 32B pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+15.7points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+7.2points ahead
- Code for real scientific research problemsSciCode+3.9points ahead
Where Apertus 8B Instruct pulls ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+1.4points ahead
Every test, side by side
All 8 tests both models report. The winning score is in its model's colour; marks a score checked independently.
ReasoningEven
- GPQA DiamondOLMo 2 32B by 7.225.632.8+7.2
- Humanity's Last ExamApertus 8B Instruct by 1.453.6+1.4
Following instructionsOLMo 2 32B
- IFBenchOLMo 2 32B by 15.722.438.1+15.7
Other results2 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.
- τ²-Bench Telecom (AA run)Apertus 8B Instruct by 11.411.40+11.4
- AA IntelligenceOLMo 2 32B by 1.24.86+1.2
Questions people ask
Which is better, Apertus 8B Instruct or OLMo 2 32B?
OLMo 2 32B wins two of the four areas where both have results: coding and following instructions. Apertus 8B Instruct wins none. They are level on reasoning and long documents.
Which is better for coding?
OLMo 2 32B. It wins 1 of the 2 coding tests both models report; Apertus 8B Instruct wins none, and 1 is a tie.
How do you compare the two?
We use the 8 benchmark tests both models have published scores on. The verdict counts the 6 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 2 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.