math benchmark
PhiBench Leaderboard
PhiBench is an internal benchmark created by Microsoft's research team to evaluate diverse skills and reasoning abilities in language models across coding and mathematics tasks.
Updated Sep 5, 2026
Models3
Model coverage3
MetricScore
EvidenceB
PhiBench Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelMI | Score74.20% | Percentile100.00% | Participants3 | EvidenceC | Evaluated |
| Rank02 | ModelMI | Score70.60% | Percentile50.00% | Participants3 | EvidenceC | Evaluated |
| Rank03 | ModelMI | Score56.20% | Percentile0.00% | Participants3 | EvidenceC | Evaluated |
PhiBench Highlights
The leading models and scores on this benchmark.
PhiBench Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for PhiBench
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is PhiBench?
What PhiBench measures and how its scores work.
PhiBench is an internal benchmark covering tasks such as debugging, extending incomplete code, explaining code snippets, identifying proof errors, and generating related problems.
It measures diverse skills and reasoning abilities in coding and mathematics, reported using the Score metric as a ratio.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- PhiBench
- Modality
- text
- Primary category
- math
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about PhiBench.
Which model scores highest on PhiBench?
Phi 4 Reasoning Plus is currently ranked first with 74.20%.
What are the top three models on PhiBench?
The current leaders are Phi 4 Reasoning Plus (74.20%), Phi 4 Reasoning (70.60%), and Phi 4 (56.20%).
Which PhiBench model has the lowest official input price?
Phi 4 Reasoning Plus has the lowest matched official input price at $0.13 input / $0.50 output per 1M tokens.
Which models are fastest among PhiBench results?
The fastest matched records are Phi 4 (33.00 tok/s via DeepInfra).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.