Microsoft
availableShows if the model has enough results for an index.Phi-4
Phi-4 is a non-reasoning model from Microsoft. 12 benchmarks count toward its score, in 5 categories.
IndexOverall score out of 100.23.3 ±8.2
CoverageShare of the index weight with results.65%
SpeedOutput tokens per second.12/s
Input / 1MUS dollars per 1M input tokens.$0.07
Output / 1MUS dollars per 1M output tokens.$0.14
ContextMaximum tokens in one request.16K
EloLMArena rating and rank.1217 (#300)
The index is a score out of 100. The ± range shows how much it can change.
24,126 votes. Elo shows what people prefer. It does not change the score.
CapabilitiesScore per category, out of 100.
Out of 100Results
12 counted| BenchmarkThe test name. | CategoryThe capability that the test measures. | ResultThe score from the publisher. | IndexThis result as a score out of 100. | RunThe settings of the run. | DateDate of the result. | Published byThe source of the result. |
|---|---|---|---|---|---|---|
| MATH level 5 | Math | 64.9% | 30.6 | — | — | Epoch AI |
| Artificial Analysis GPQA Diamond | Knowledge | 57.5% | 27.5 | — | — | Artificial Analysis |
| GPQA diamond | Knowledge | 56.1% | 30.0 | — | — | Epoch AI |
| Artificial Analysis IFBench | Instruction | 23.5% | 13.0 | — | — | Artificial Analysis |
| Artificial Analysis Omniscience Accuracy | Knowledge | 14.1% | 31.3 | — | — | Artificial Analysis |
| OTIS Mock AIME 2024-2025 | Math | 13.8% | 19.3 | — | — | Epoch AI |
| Artificial Analysis Intelligence Index | Knowledge | 5.9% | 29.8 | — | — | Artificial Analysis |
| Artificial Analysis Humanity's Last Exam | Knowledge | 3.8% | 28.7 | — | — | Artificial Analysis |
| Chess Puzzles | Reasoning | 1.0% | 24.8 | — | — | Epoch AI |
| τ²-Bench Tool-Agent-User Evaluation | Agentic | 0.0% | 5.0 | — | — | Victor Barres et al. |
| Artificial Analysis Long Context Reasoning | Reasoning | 0.0% | 8.3 | — | — | Artificial Analysis |
| Critical Physics Tasks | Reasoning | 0.0% | 38.4 | — | — | Artificial Analysis |
12 benchmarks count, from 12 of 12 results. A grey row does not count. Too few models took that benchmark.