Qwen2.5-72B
Qwen2.5-72B (Alibaba) was released on Sep 19, 2024. In the Epoch AI capability ranking (ECI), it ranks #166 out of 267 models, with 129.00. Against models of the same generation, its best result is on ARC (AI2): 94.5%, #2 of 22.
Where Qwen2.5-72B stands out and where it falls behind, by the numbers
Qwen2.5-72B has published results on 21 tests compiled by Epoch AI. The yardstick here is its generation: its position among the models released between Dec 24, 2023 and Dec 18, 2024 that took the same test. Against the whole archive, which includes 2023 models, a recent release lands near the top of almost everything, and that says nothing.
Where it stands out
Tests where it sits in the top third of its own generation.
Where it falls behind
Tests where it sits in the bottom half of its own generation.
See all 21 tests, one by one
| Test | Score | In its generation | Overall | Test record |
|---|---|---|---|---|
| GPQA Diamond | 49.1% | #10 / 35 | #130 / 173 | 96.4% Gemini 4 Argon |
| AIME 2024 and 2025 (OTIS) | 8.1% | #10 / 30 | #137 / 168 | 100.0% GPT-5.5 |
| Dtbench | 62.9% | #5 / 28 | #103 / 148 | 98.9% Claude Opus 5.5 |
| Lmca | 13.42 | #6 / 14 | #105 / 121 | 68.23 Claude Opus 5.5 |
| WeirdML | 16.0% | #12 / 20 | #101 / 112 | 93.6% GPT-6 Astra |
| MMLU | 85.3% | #5 / 48 | #7 / 92 | 88.1% GPT-4o (May 2024) |
| MATH level 5 | 63.2% | #6 / 33 | #35 / 73 | 98.1% GPT-5 |
| ARC (AI2) | 94.5% | #2 / 22 | #3 / 65 | 95.3% DeepSeek-V3 (Mar 2025) |
| ForecastBench | 57.50 | #9 / 14 | #49 / 62 | 62.50 o3 |
| WinoGrande | 82.3% | #5 / 27 | #9 / 61 | 89.2% Llama 3.1-405B |
| HellaSwag | 84.8% | #3 / 17 | #10 / 51 | 95.3% GPT-4 (Mar 2023) |
| PIQA | 82.6% | #7 / 11 | #15 / 43 | 88.7% GPT-4o mini |
| BIG-Bench Hard | 79.8% | #4 / 11 | #5 / 41 | 89.2% Gemini 1.5 Pro (Sept 2024) |
| BALROG | 16.2% | #10 / 15 | #27 / 32 | 68.3% GPT-6 Astra |
| TriviaQA | 71.9% | #5 / 9 | #24 / 32 | 87.6% Llama 2-70B |
| METR: time horizon | 5.17 | #5 / 7 | #27 / 29 | 719 Claude Opus 4.6 |
| GeoBench | 3,448 | #4 / 8 | #15 / 24 | 4,333 Gemini 3 Flash |
| The Agent Company | 11.8% | #5 / 8 | #11 / 14 | 52.4% DeepSeek-V3.2-Exp |
| Os World | 5.00 | #1 / 1 | #8 / 8 | 72.10 Claude Sonnet 4.6 |
| Spatialviz Bench | 33.3% | #2 / 2 | #6 / 7 | 44.7% Gemini 2.5 Pro (Jun 2025) |
| Video Mme | 73.5% | #2 / 6 | #2 / 6 | 75.0% Gemini 1.5 Pro (Sept 2024) |
For each test, the model's best effort level counts, as in the ranking. Source of the scores: Epoch AI, under CC BY 4.0.
What to compare Qwen2.5-72B with
The models from other labs closest in capability and released in the same period, at most two per lab. Click any of them to see its page.
| Model | Lab | ECI | ELO | LiveBench | Release |
|---|---|---|---|---|---|
| Llama 4 Scout | Meta AI | 129.65 | n/a | n/a | Apr 5, 2025 |
| GPT-4.1 nano | OpenAI | 129.63 | n/a | n/a | Apr 14, 2025 |
| Gemini 1.5 Flash (Sep 2024) | Google DeepMind | 129.37 | n/a | n/a | Sep 24, 2024 |
| GPT-4o (May 2024) | OpenAI | 129.05 | n/a | n/a | May 13, 2024 |
| Qwen2.5-72B | Alibaba | 129.00 | n/a | n/a | Sep 19, 2024 |
| Llama 3.1-405B | Meta AI | 128.76 | n/a | n/a | Jul 23, 2024 |
| Mistral Large 2 (Nov 2024) | Mistral AI | 128.53 | n/a | n/a | Nov 18, 2024 |
Qwen2.5-72B or GPT-4o (May 2024)?
GPT-4o (May 2024) is the model from another lab closest to Qwen2.5-72B in capability. Across the 13 tests both took, this is the score:
| Test | Qwen2.5-72B | GPT-4o (May 2024) |
|---|---|---|
| BALROG | 16.2% | 32.3% |
| METR: time horizon | 5.17 | 9.17 |
| WeirdML | 16.0% | 25.1% |
| The Agent Company | 11.8% | 16.7% |
| AIME 2024 and 2025 (OTIS) | 8.1% | 6.4% |
| Lmca | 13.42 | 16.59 |
| MATH level 5 | 63.2% | 53.3% |
| GeoBench | 3,448 | 3,696 |
| MMLU | 85.3% | 88.1% |
| Dtbench | 62.9% | 64.5% |
Unweighted score: each test counts one point, whether math or coding. The dot marks who did better on each row.
Other models from Alibaba
The models from the same lab closest in time, newest first.
| Model | Release | ECI | ELO | LiveBench | Access |
|---|---|---|---|---|---|
| Qwen3-8B | Apr 28, 2025 | 136.17 | n/a | n/a | Open weights (unrestricted) |
| QwQ-32B | Mar 5, 2025 | 137.58 | 1,378 | n/a | Open weights (unrestricted) |
| Qwen2.5-Max | Jan 25, 2025 | 132.53 | 1,396 | n/a | API access |
| Qwen2.5-72B | Sep 19, 2024 | 129.00 | n/a | n/a | Open weights (unrestricted) |
| Qwen2.5-7B | Sep 19, 2024 | 118.53 | n/a | n/a | Open weights (unrestricted) |
| Qwen2.5-Coder-32B | Sep 18, 2024 | 119.52 | n/a | n/a | Open weights (unrestricted) |
| Qwen2.5-Coder (7B) | Sep 18, 2024 | 113.20 | n/a | n/a | Open weights (unrestricted) |
| Qwen2.5-32B | Sep 17, 2024 | 128.53 | n/a | n/a | Open weights (unrestricted) |
| Qwen2-72B | Jun 7, 2024 | 125.28 | n/a | n/a | Open weights (unrestricted) |
Have you used Qwen2.5-72B? Rate it
Users' opinion, not measurement: the rating is kept apart from the ranking and does not move anyone's position. One vote per browser, no sign-up, and you can change your rating later.
What people ask about Qwen2.5-72B
Where does Qwen2.5-72B rank among AI models?
In the ROO3 AI Benchmark, ordered by the Epoch AI ECI, Qwen2.5-72B ranks #166 out of 267 models, with an ECI of 129.00. The ECI is a composite index that combines several hard benchmarks into a single scale. The position changes when a new model comes out, and the page is refreshed six times a day.
Is Qwen2.5-72B better than GPT-4o (May 2024)?
Across the 13 tests both models took and Epoch AI compiled, Qwen2.5-72B comes out ahead on 3, behind on 10 and tied on 0. It is a test-by-test count with no weighting: a win in math counts the same as a win in coding. To choose between the two, look at the tests that resemble your own use.
What is Qwen2.5-72B best at?
Against models of the same generation (released up to 9 months before and up to 3 months after it), the best relative results of Qwen2.5-72B are: ARC (AI2) (#2 of 22); MMLU (#5 of 48); HellaSwag (#3 of 17). The yardstick is its position among the models of that period that took the same test.
Where do the numbers on this page come from?
From the same three sources as the ranking, all openly licensed: Epoch AI (ECI and individual tests, CC BY 4.0), Arena (human-preference ELO, CC BY 4.0) and LiveBench (overall average, CC BY-SA 4.0). ROO3 does not measure any model. When the page shows a figure published by the maker itself, it appears in the written analysis, labeled as the maker's, and never enters the ranking.
Where each number comes from
Epoch AI, "AI Benchmarking Hub". Published at epoch.ai. Data under the CC BY 4.0 licence, reordered by roo3.co.
The Arena (LMArena) leaderboard, from the public lmarena-ai/leaderboard-dataset on Hugging Face, under the CC BY 4.0 licence. We are not affiliated with the Arena.
LiveBench, by the LiveBench team, published at livebench.ai under the CC BY-SA 4.0 license. The overall average is recomputed by roo3.co with the same formula as the website (average of the categories), and those numbers stay under CC BY-SA 4.0. We are not affiliated with LiveBench.
← See the full ranking