agents benchmark
DeepSWE 1.1
DeepSWE 1.1 evaluates software engineering agents using the mini-swe-agent harness.
Updated Aug 7, 2026
Published models19
Registry coverage19
MetricScore
EvidenceB
DeepSWE 1.1 leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | OP | 73.0% | 100.0% | 19 | B | |
| 2 | AN | 70.0% | 94.4% | 19 | B | |
| 3 | OP | 70.0% | 88.9% | 19 | B | |
| 4 | MA | 69.0% | 83.3% | 19 | B | |
| 5 | AN | 68.8% | 77.8% | 19 | C | |
| 6 | OP | 67.0% | 72.2% | 19 | B | |
| 7 | OP | 67.0% | 66.7% | 19 | B | |
| 8 | AN | 59.0% | 61.1% | 19 | B | |
| 9 | AC | 56.6% | 55.6% | 19 | C | |
| 10 | AN | 54.0% | 50.0% | 19 | B | |
| 11 | XA | 54.0% | 44.4% | 19 | B | |
| 12 | ME | 53.0% | 38.9% | 19 | B | |
| 13 | OP | 52.0% | 33.3% | 19 | B | |
| 14 | GO | 49.0% | 27.8% | 19 | B | |
| 15 | ZA | 44.0% | 22.2% | 19 | B | |
| 16 | GO | 37.0% | 16.7% | 19 | B | |
| 17 | MA | 31.0% | 11.1% | 19 | B | |
| 18 | AN | 30.0% | 5.6% | 19 | B | |
| 19 | GO | 12.0% | 0.0% | 19 | B |
Score distribution
Top published rows on the benchmark's original scale.
DeepSWE 1.1 highlights
The top published results on this benchmark's own scale.
What is DeepSWE 1.1?
Definition and scoring fields from the benchmark registry.
DeepSWE 1.1 evaluates software engineering agents using the mini-swe-agent harness.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- DeepSWE 1.1
- Modality
- text
- Primary category
- agents
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- deepswe-1.1|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about DeepSWE 1.1.
Which model scores highest on DeepSWE 1.1?
GPT-5.6 Sol is currently ranked first with 73.0%.
What does DeepSWE 1.1 measure?
DeepSWE 1.1 evaluates software engineering agents using the mini-swe-agent harness.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
19 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.