agents benchmark
DeepSWE Leaderboard
DeepSWE is a software engineering agent benchmark evaluated with the mini-swe-agent harness, where each task is solved in an isolated container with no internet access. It measures an agent's ability to autonomously resolve real-world coding issues end to end.
Updated Aug 17, 2026
Models11
Model coverage11
MetricScore
EvidenceB
DeepSWE Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelOP | Score72.7% | Percentile100.0% | Participants11 | EvidenceC | Evaluated |
| Rank02 | ModelOP | Score69.6% | Percentile90.0% | Participants11 | EvidenceC | Evaluated |
| Rank03 | ModelMA | Score67.5% | Percentile80.0% | Participants11 | EvidenceC | Evaluated |
| Rank04 | ModelOP | Score67.2% | Percentile70.0% | Participants11 | EvidenceC | Evaluated |
| Rank05 | ModelDE | Score62.7% | Percentile60.0% | Participants11 | EvidenceC | Evaluated |
| Rank06 | ModelDE | Score54.4% | Percentile50.0% | Participants11 | EvidenceC | Evaluated |
| Rank07 | ModelXA | Score53.0% | Percentile40.0% | Participants11 | EvidenceC | Evaluated |
| Rank08 | ModelZA | Score46.2% | Percentile30.0% | Participants11 | EvidenceC | Evaluated |
| Rank09 | ModelBY | Score32.7% | Percentile20.0% | Participants11 | EvidenceC | Evaluated |
| Rank10 | ModelTE | Score28.0% | Percentile10.0% | Participants11 | EvidenceC | Evaluated |
| Rank11 | ModelBY | Score23.0% | Percentile0.0% | Participants11 | EvidenceC | Evaluated |
DeepSWE Highlights
The leading models and scores on this benchmark.
DeepSWE Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for DeepSWE
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis deepswe 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.
- 01OPGPT-5.6 SolOpenAI
- Score
- 72.7%
- Price
- $5.0 input / $30 output per 1M tokens
- Speed
- Up to 27 tok/s via OpenAI
Strengths
- Ranks #1 of 11 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures DeepSWE, not total model capability
- 02OPGPT-5.6 TerraOpenAI
- Score
- 69.6%
- Price
- $2.0 input / $12 output per 1M tokens
- Speed
- Up to 35 tok/s via OpenAI
Strengths
- Ranks #2 of 11 compared models
- 90th percentile on this benchmark
- C evidence result
Considerations
- This result measures DeepSWE, not total model capability
- 03MAKimi K3Moonshot AI
- Score
- 67.5%
- Price
- $3.0 input / $15 output per 1M tokens
- Speed
- Up to 26 tok/s via Fireworks
Strengths
- Ranks #3 of 11 compared models
- 80th percentile on this benchmark
- C evidence result
Considerations
- This result measures DeepSWE, not total model capability
- 04OPGPT-5.6 LunaOpenAI
- Score
- 67.2%
- Price
- $0.20 input / $1.2 output per 1M tokens
- Speed
- Up to 49 tok/s via OpenAI
Strengths
- Ranks #4 of 11 compared models
- 70th percentile on this benchmark
- C evidence result
Considerations
- This result measures DeepSWE, not total model capability
- 05DEDeepSeek-V4-Pro-0813DeepSeek
- Score
- 62.7%
- Price
- $0.43 input / $0.87 output per 1M tokens
- Speed
- Up to 33 tok/s via DeepSeek
Strengths
- Ranks #5 of 11 compared models
- 60th percentile on this benchmark
- C evidence result
Considerations
- This result measures DeepSWE, not total model capability
Selection summary
Best AI Models for DeepSWE
GPT-5.6 Sol currently leads DeepSWE with 72.7%. 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 DeepSWE?
What DeepSWE measures and how its scores work.
DeepSWE is a software engineering agent benchmark evaluated with the mini-swe-agent harness, where each task is solved in an isolated container with no internet access. It measures an agent's ability to autonomously resolve real-world coding issues end to end.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- DeepSWE
- Modality
- text
- Primary category
- agents
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- deepswe|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about DeepSWE.
Which model scores highest on DeepSWE?
GPT-5.6 Sol is currently ranked first with 72.7%.
What does DeepSWE measure?
DeepSWE is a software engineering agent benchmark evaluated with the mini-swe-agent harness, where each task is solved in an isolated container with no internet access. It measures an agent's ability to autonomously resolve real-world coding issues end to end.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
11 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.