long context benchmark
RepoQA Leaderboard
RepoQA is a benchmark for evaluating long-context code understanding capabilities of Large Language Models through the Searching Needle Function (SNF) task, where LLMs must locate specific functions in code repositories using natural language descriptions. The benchmark contains 500 code search tasks spanning 50 repositories across 5 modern programming languages (Python, Java, TypeScript, C++, and Rust), tested on 26 general and code-specific LLMs to assess their ability to comprehend and navigate code repositories.
Updated Aug 17, 2026
Models2
Model coverage2
MetricScore
EvidenceB
RepoQA Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelMI | Score85.0% | Percentile100.0% | Participants2 | EvidenceC | Evaluated |
| Rank02 | ModelMI | Score77.0% | Percentile0.0% | Participants2 | EvidenceC | Evaluated |
RepoQA Highlights
The leading models and scores on this benchmark.
RepoQA Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for RepoQA
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis repoqa AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01MIPhi-3.5-MoE-instructMicrosoft
- Score
- 85.0%
Strengths
- Ranks #1 of 2 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures RepoQA, not total model capability
- 02MIPhi-3.5-mini-instructMicrosoft
- Score
- 77.0%
- Speed
- Up to 23 tok/s via Azure
Strengths
- Ranks #2 of 2 compared models
- 0th percentile on this benchmark
- C evidence result
Considerations
- This result measures RepoQA, not total model capability
Selection summary
Best AI Models for RepoQA
Phi-3.5-MoE-instruct currently leads RepoQA with 85.0%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is RepoQA?
What RepoQA measures and how its scores work.
RepoQA is a benchmark for evaluating long-context code understanding capabilities of Large Language Models through the Searching Needle Function (SNF) task, where LLMs must locate specific functions in code repositories using natural language descriptions. The benchmark contains 500 code search tasks spanning 50 repositories across 5 modern programming languages (Python, Java, TypeScript, C++, and Rust), tested on 26 general and code-specific LLMs to assess their ability to comprehend and navigate code repositories.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- RepoQA
- Modality
- text
- Primary category
- long context
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- repoqa|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about RepoQA.
Which model scores highest on RepoQA?
Phi-3.5-MoE-instruct is currently ranked first with 85.0%.
What does RepoQA measure?
RepoQA is a benchmark for evaluating long-context code understanding capabilities of Large Language Models through the Searching Needle Function (SNF) task, where LLMs must locate specific functions in code repositories using natural language descriptions. The benchmark contains 500 code search tasks spanning 50 repositories across 5 modern programming languages (Python, Java, TypeScript, C++, and Rust), tested on 26 general and code-specific LLMs to assess their ability to comprehend and navigate code repositories.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
2 model results are currently shown.
Does this benchmark affect the overall score?
No. This benchmark is shown for reference but does not contribute to the overall score.