agents benchmark
SWE-Marathon Leaderboard
SWE-Marathon is an ultra-long-horizon software engineering benchmark covering tasks such as building compilers, optimizing kernels, and developing production-grade services. It measures whether agents can sustain quality across extremely long engineering trajectories.
Updated Aug 17, 2026
Models4
Model coverage4
MetricScore
EvidenceB
SWE-Marathon Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelZA | Score42.5% | Percentile100.0% | Participants4 | EvidenceC | Evaluated |
| Rank02 | ModelMA | Score42.0% | Percentile66.7% | Participants4 | EvidenceC | Evaluated |
| Rank03 | ModelXA | Score29.0% | Percentile33.3% | Participants4 | EvidenceC | Evaluated |
| Rank04 | ModelZA | Score13.0% | Percentile0.0% | Participants4 | EvidenceC | Evaluated |
SWE-Marathon Highlights
The leading models and scores on this benchmark.
SWE-Marathon Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for SWE-Marathon
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis swe-marathon 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.
- 01ZAGLM-5.3Zhipu AI
- Score
- 42.5%
Strengths
- Ranks #1 of 4 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures SWE-Marathon, not total model capability
- 02MAKimi K3Moonshot AI
- Score
- 42.0%
- Price
- $3.0 input / $15 output per 1M tokens
- Speed
- Up to 26 tok/s via Fireworks
Strengths
- Ranks #2 of 4 compared models
- 67th percentile on this benchmark
- C evidence result
Considerations
- This result measures SWE-Marathon, not total model capability
- 03XAGrok 4.5xAI
- Score
- 29.0%
- Price
- $2.0 input / $6.0 output per 1M tokens
- Speed
- Up to 4.3 tok/s via xAI
Strengths
- Ranks #3 of 4 compared models
- 33th percentile on this benchmark
- C evidence result
Considerations
- This result measures SWE-Marathon, not total model capability
- 04ZAGLM-5.2Zhipu AI
- Score
- 13.0%
- Price
- $1.4 input / $4.4 output per 1M tokens
Strengths
- Ranks #4 of 4 compared models
- 0th percentile on this benchmark
- C evidence result
Considerations
- This result measures SWE-Marathon, not total model capability
Selection summary
Best AI Models for SWE-Marathon
GLM-5.3 currently leads SWE-Marathon with 42.5%. 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 SWE-Marathon?
What SWE-Marathon measures and how its scores work.
SWE-Marathon is an ultra-long-horizon software engineering benchmark covering tasks such as building compilers, optimizing kernels, and developing production-grade services. It measures whether agents can sustain quality across extremely long engineering trajectories.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- SWE-Marathon
- Modality
- text
- Primary category
- agents
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- swe-marathon|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about SWE-Marathon.
Which model scores highest on SWE-Marathon?
GLM-5.3 is currently ranked first with 42.5%.
What does SWE-Marathon measure?
SWE-Marathon is an ultra-long-horizon software engineering benchmark covering tasks such as building compilers, optimizing kernels, and developing production-grade services. It measures whether agents can sustain quality across extremely long engineering trajectories.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
4 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.