How the models measure up.
A live leaderboard of AI models on public tests of reasoning, maths, coding and knowledge, and where Qorra stands among them.
Go deeper: Meet Qorra · Model Fab
- 1GPT-6 Astra98.5
- 2GPT-6.1 Sol97.9
- 3GPT 5.5 Pre Release95.1
- 4Claude Opus 5.594.2
- 5GPT-6.1 Terra93.0
Every score, in one place.
QorraaskFinz | — | — | — | — | — | — | — | — | 5.3* | — | — |
GPT-6 Astra | 98.5 | — | 95.8 | 93.7 | 97.6 | 100.0 | — | — | 75.6 | 76.0 | 84.0 |
GPT-6.1 Sol | 97.9 | — | 95.4 | 93.7 | 100.0 | 100.0 | — | — | 73.9 | 68.0 | 80.0 |
GPT 5.5 Pre Release | 95.1 | — | 94.0 | — | — | 100.0 | — | 80.6 | — | — | — |
Claude Opus 5.5 | 94.2 | — | 90.6 | 91.2 | 95.0 | 100.0 | — | — | 72.2 | 67.0 | 71.0 |
GPT-6.1 Terra | 93.0 | — | 94.3 | 90.9 | 92.7 | 100.0 | — | — | 57.3 | — | 74.0 |
Gemini 4 Argon | 92.7 | — | 96.4 | 86.0 | 78.0 | 100.0 | — | — | 76.0 | 59.0 | 64.0 |
GPT-5.6 Sol | 92.6 | — | 93.5 | 89.1 | 82.9 | 100.0 | — | — | 69.7 | 68.0 | 58.0 |
GPT-6 Sol | 90.4 | — | 94.3 | 89.8 | 90.0 | 100.0 | — | — | 60.7 | 56.0 | 56.0 |
Claude Fable 5.1 | 90.1 | — | — | 90.2 | 87.8 | 100.0 | — | — | 70.8 | 48.0 | 58.0 |
GPT-5.5 Pro | 89.8 | — | — | 87.7 | 78.0 | — | — | — | — | 62.0 | — |
Claude Sonnet 5.5 | 87.6 | — | 95.6 | 88.8 | 80.5 | 100.0 | — | — | 46.5 | 59.0 | 65.0 |
Claude Fable 5 | 85.6 | — | 85.9 | 87.0 | 90.2 | 100.0 | — | — | 70.7 | 47.0 | 52.0 |
Muse Wonder | 84.3 | — | — | 82.1 | 70.7 | 99.4 | — | — | — | — | 51.0 |
Claude Opus 5 | 84.1 | — | 93.9 | 85.6 | 73.2 | 98.9 | — | — | 59.9 | 43.0 | 59.0 |
GPT-5.5 | 82.3 | — | 90.7 | 85.3 | 72.5 | 84.4 | — | — | 63.0 | 63.0 | 56.0 |
Gemini 3.7 Flash | 82.0 | — | 94.8 | 71.6 | 36.6 | 97.2 | — | — | 69.2 | 66.0 | 37.0 |
GPT-5.4 Pro | 80.1 | — | 94.6 | 82.5 | 58.5 | — | — | — | 46.3 | 59.0 | — |
Claude Opus 4.8 | 79.6 | — | 91.0 | 80.0 | 56.1 | 98.3 | — | — | 53.0 | 52.0 | 36.0 |
GPT-5.6 Terra | 78.0 | — | 93.3 | 86.0 | 70.7 | 99.7 | — | — | 43.2 | 47.0 | 35.0 |
Gemini 3.8 Flash | 77.7 | — | 95.4 | 68.4 | 22.0 | 98.9 | — | — | 69.7 | 58.0 | 47.0 |
GPT-5.2 Pro | 76.5 | — | — | 74.0 | 46.0 | — | — | — | — | 60.0 | — |
GPT-5.4 | 75.8 | — | 93.3 | 78.6 | 49.0 | 97.8 | — | 76.9 | 45.1 | 48.0 | 37.0 |
Qwen3.8 Max | 75.1 | — | 92.7 | 74.7 | 46.3 | 99.4 | — | — | 45.8 | 46.0 | 38.0 |
DeepSeek V4.1 Flash 1017 | 74.2 | 763B | 93.2 | 71.2 | 46.3 | 99.4 | — | — | 42.5 | 46.0 | 49.0 |
Gemini 3.1 Pro Preview | 73.8 | — | 94.4 | 59.6 | 26.8 | 95.6 | — | — | 73.5 | 51.0 | 34.0 |
Muse Spark Open | 72.1 | 822B | 88.6 | 69.8 | 31.7 | 98.1 | — | — | 56.6 | — | — |
Gemini 3.5 Flash | 71.9 | — | 92.8 | 62.8 | 26.8 | 95.6 | — | 79.3 | 66.2 | 45.0 | 32.0 |
Claude Opus 4.7 | 71.7 | — | 90.2 | 70.2 | 31.7 | 97.8 | — | 83.5 | 51.7 | 41.0 | 28.0 |
DeepSeek V4 Pro 0813 | 70.7 | 1.7T | 91.7 | 64.6 | 26.8 | 98.6 | — | — | 52.9 | 46.0 | 43.0 |
Kimi K3 | 69.3 | — | 93.1 | 72.2 | 39.0 | 97.2 | — | — | 50.6 | 37.0 | 26.0 |
Its own test runs.
- MMLU-Pro45.5Qorra alone · 200 questions · 2026-10-30
- Global-MMLU-Lite62.5Qorra alone · 144 questions · 2026-10-30
- MGSM19.7Qorra alone · 132 questions · 2026-10-30
- SimpleQA Verified5.3Qorra alone · 150 questions · 2026-10-30
- SimpleQA Verified5.3With askFinz Search · 150 questions · 2026-10-30
- FRAMES (Google)8.0Qorra alone · 150 questions · 2026-10-30
- FRAMES (Google)6.0With askFinz Search · 150 questions · 2026-10-30
* Run by askFinz, not by the group behind the table. Marking can differ between the two, so read a starred score as a guide rather than an exact match.
Made for different work.
Asked most.
Where do these scores come from?
The table uses results published by independent testers, and it refreshes as new results come out. The sources are credited at the foot of this page.
What is the board score?
It is a model's average standing across the tests it took, on a scale of 0 to 100. A model needs at least three tests to get one, and it says how a model ranks against others, not how many questions it got right.
Why does Qorra show a star?
A star means askFinz ran the test itself. Marking can differ from the group behind the table, so read a starred score as a guide rather than an exact match.
Why is a cell empty?
An empty cell means there is no result for that model on that test. We never estimate a missing score.
Why are Sonar, Tirra and Unira not in the table?
They do specialist jobs that general model tests do not measure, so there is nothing fair to put in a row.
Each score is a model's best run. A dash means no result. Size is a model's parameter count, which closed models do not publish.
Scores: Epoch AI, “Capabilities & benchmarking”, licensed under CC BY 4.0; scores were rescaled to 0-100 and combined into the board score. Prices and context: models.dev. Size: Hugging Face, or the model's published name.
Be there when Qorra arrives.
askFinz is in private access, and new accounts join in waves. Request access and you will meet Qorra inside askFinz as it rolls out.