Phi-4
Phi-4 (Microsoft Research) was released on Dec 12, 2024. In the Epoch AI capability ranking (ECI), it ranks #160 out of 267 models, with 130.43. On Arena, which measures human preference in blind comparisons, it has 1,231 ELO points (#299). Against models of the same generation, its best result is on MMLU: 84.8%, #6 of 38.
Where Phi-4 stands out and where it falls behind, by the numbers
Phi-4 has published results on 7 tests compiled by Epoch AI. The yardstick here is its generation: its position among the models released between Mar 17, 2024 and Mar 12, 2025 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 7 tests, one by one
| Test | Score | In its generation | Overall | Test record |
|---|---|---|---|---|
| GPQA Diamond | 56.1% | #12 / 44 | #121 / 173 | 96.4% Gemini 4 Argon |
| AIME 2024 and 2025 (OTIS) | 13.8% | #15 / 38 | #132 / 168 | 100.0% GPT-5.5 |
| MMLU | 84.8% | #6 / 38 | #8 / 92 | 88.1% GPT-4o (May 2024) |
| MATH level 5 | 64.9% | #12 / 38 | #33 / 73 | 98.1% GPT-5 |
| Creative writing (Lech Mazur) | 6.26 | #15 / 17 | #35 / 38 | 8.60 GPT-5 |
| BALROG | 11.6% | #14 / 16 | #30 / 32 | 68.3% GPT-6 Astra |
| LiveBench | 41.61 | #20 / 26 | #26 / 32 | 82.35 Gemini 2.5 Pro (Jun 2025) |
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 Phi-4 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 |
|---|---|---|---|---|---|
| Mistral Small 3.2 | Mistral AI | 131.75 | n/a | n/a | Jun 20, 2025 |
| Grok-2 (Dec 2024) | xAI | 130.49 | n/a | n/a | Dec 12, 2024 |
| Phi-4 | Microsoft Research | 130.43 | 1,231 | n/a | Dec 12, 2024 |
| Gemma 3 27B | Google DeepMind | 130.05 | n/a | n/a | Mar 12, 2025 |
| 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 |
Phi-4 or Grok-2 (Dec 2024)?
Grok-2 (Dec 2024) is the model from another lab closest to Phi-4 in capability. Across the 5 tests both took, this is the score:
| Test | Phi-4 | Grok-2 (Dec 2024) |
|---|---|---|
| LiveBench | 41.61 | 54.30 |
| AIME 2024 and 2025 (OTIS) | 13.8% | 11.5% |
| GPQA Diamond | 56.1% | 53.8% |
| MATH level 5 | 64.9% | 63.5% |
| Creative writing (Lech Mazur) | 6.26 | 6.36 |
Unweighted score: each test counts one point, whether math or coding. The dot marks who did better on each row.
Have you used Phi-4? 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 Phi-4
Where does Phi-4 rank among AI models?
In the ROO3 AI Benchmark, ordered by the Epoch AI ECI, Phi-4 ranks #160 out of 267 models, with an ECI of 130.43. 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 Phi-4 better than Grok-2 (Dec 2024)?
Across the 5 tests both models took and Epoch AI compiled, Phi-4 comes out ahead on 3, behind on 2 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 Phi-4 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 Phi-4 are: MMLU (#6 of 38); GPQA Diamond (#12 of 44); MATH level 5 (#12 of 38). 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