reasoning benchmark
SWE-Bench Pro Leaderboard
SWE-Bench Pro is an advanced version of SWE-Bench that evaluates language models on complex, real-world software engineering tasks requiring extended reasoning and multi-step problem solving.
Updated Aug 17, 2026
Models50
Model coverage50
MetricScore
EvidenceB
SWE-Bench Pro Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAN | Score80.0% | Percentile100.0% | Participants50 | EvidenceC | Evaluated |
| Rank02 | ModelAN | Score77.8% | Percentile98.0% | Participants50 | EvidenceC | Evaluated |
| Rank03 | ModelAN | Score69.2% | Percentile95.9% | Participants50 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score67.7% | Percentile93.9% | Participants50 | EvidenceC | Evaluated |
| Rank05 | ModelXA | Score64.7% | Percentile91.8% | Participants50 | EvidenceC | Evaluated |
| Rank06 | ModelOP | Score64.6% | Percentile89.8% | Participants50 | EvidenceC | Evaluated |
| Rank07 | ModelAN | Score64.3% | Percentile87.8% | Participants50 | EvidenceC | Evaluated |
| Rank08 | ModelOP | Score63.4% | Percentile85.7% | Participants50 | EvidenceC | Evaluated |
| Rank09 | ModelAN | Score63.2% | Percentile83.7% | Participants50 | EvidenceC | Evaluated |
| Rank10 | ModelOP | Score62.7% | Percentile81.6% | Participants50 | EvidenceC | Evaluated |
| Rank11 | ModelZA | Score62.1% | Percentile79.6% | Participants50 | EvidenceC | Evaluated |
| Rank12 | ModelAC | Score61.7% | Percentile77.5% | Participants50 | EvidenceC | Evaluated |
| Rank13 | ModelME | Score61.5% | Percentile75.5% | Participants50 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score60.6% | Percentile73.5% | Participants50 | EvidenceC | Evaluated |
| Rank15 | ModelPO | Score59.4% | Percentile71.4% | Participants50 | EvidenceC | Evaluated |
| Rank16 | ModelMI | Score59.0% | Percentile69.4% | Participants50 | EvidenceC | Evaluated |
| Rank17 | ModelGO | Score58.7% | Percentile67.3% | Participants50 | EvidenceC | Evaluated |
| Rank18 | ModelOP | Score58.6% | Percentile65.3% | Participants50 | EvidenceC | Evaluated |
| Rank19 | ModelMA | Score58.6% | Percentile63.3% | Participants50 | EvidenceC | Evaluated |
| Rank20 | ModelZA | Score58.4% | Percentile61.2% | Participants50 | EvidenceC | Evaluated |
| Rank21 | ModelTE | Score57.9% | Percentile59.2% | Participants50 | EvidenceC | Evaluated |
| Rank22 | ModelOP | Score57.7% | Percentile57.1% | Participants50 | EvidenceC | Evaluated |
| Rank23 | ModelAC | Score57.6% | Percentile55.1% | Participants50 | EvidenceC | Evaluated |
| Rank24 | ModelBY | Score57.5% | Percentile53.1% | Participants50 | EvidenceC | Evaluated |
| Rank25 | ModelXI | Score57.2% | Percentile51.0% | Participants50 | EvidenceC | Evaluated |
| Rank26 | ModelBY | Score57.0% | Percentile49.0% | Participants50 | EvidenceC | Evaluated |
| Rank27 | ModelOP | Score56.8% | Percentile46.9% | Participants50 | EvidenceC | Evaluated |
| Rank28 | ModelAC | Score56.6% | Percentile44.9% | Participants50 | EvidenceC | Evaluated |
| Rank29 | ModelOP | Score56.4% | Percentile42.9% | Participants50 | EvidenceC | Evaluated |
| Rank30 | ModelMI | Score56.2% | Percentile40.8% | Participants50 | EvidenceC | Evaluated |
| Rank31 | ModelXI | Score56.1% | Percentile38.8% | Participants50 | EvidenceC | Evaluated |
| Rank32 | ModelTM | Score55.9% | Percentile36.7% | Participants50 | EvidenceC | Evaluated |
| Rank33 | ModelDE | Score55.4% | Percentile34.7% | Participants50 | EvidenceC | Evaluated |
| Rank34 | ModelMI | Score55.4% | Percentile32.6% | Participants50 | EvidenceC | Evaluated |
| Rank35 | ModelGO | Score55.1% | Percentile30.6% | Participants50 | EvidenceC | Evaluated |
| Rank36 | ModelOP | Score54.4% | Percentile28.6% | Participants50 | EvidenceC | Evaluated |
| Rank37 | ModelGO | Score54.2% | Percentile26.5% | Participants50 | EvidenceC | Evaluated |
| Rank38 | ModelGO | Score54.2% | Percentile24.5% | Participants50 | EvidenceC | Evaluated |
| Rank39 | ModelAC | Score53.5% | Percentile22.4% | Participants50 | EvidenceC | Evaluated |
| Rank40 | ModelMI | Score52.8% | Percentile20.4% | Participants50 | EvidenceC | Evaluated |
| Rank41 | ModelDE | Score52.6% | Percentile18.4% | Participants50 | EvidenceC | Evaluated |
| Rank42 | ModelOP | Score52.4% | Percentile16.3% | Participants50 | EvidenceC | Evaluated |
| Rank43 | ModelME | Score52.4% | Percentile14.3% | Participants50 | EvidenceC | Evaluated |
| Rank44 | ModelDE | Score52.3% | Percentile12.2% | Participants50 | EvidenceC | Evaluated |
| Rank45 | ModelMI | Score51.2% | Percentile10.2% | Participants50 | EvidenceC | Evaluated |
| Rank46 | ModelME | Score51.2% | Percentile8.2% | Participants50 | EvidenceC | Evaluated |
| Rank47 | ModelMA | Score50.7% | Percentile6.1% | Participants50 | EvidenceC | Evaluated |
| Rank48 | ModelAC | Score49.5% | Percentile4.1% | Participants50 | EvidenceC | Evaluated |
| Rank49 | ModelPO | Score47.6% | Percentile2.0% | Participants50 | EvidenceC | Evaluated |
| Rank50 | ModelCO | Score40.2% | Percentile0.0% | Participants50 | EvidenceC | Evaluated |
SWE-Bench Pro Highlights
The leading models and scores on this benchmark.
SWE-Bench Pro Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for SWE-Bench Pro
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis swe-bench pro 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.
- 01ANClaude Fable 5Anthropic
- Score
- 80.0%
- Price
- $10 input / $50 output per 1M tokens
- Speed
- Up to 9.1 tok/s via Anthropic
Strengths
- Ranks #1 of 50 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures SWE-Bench Pro, not total model capability
- 02ANClaude Mythos PreviewAnthropic
- Score
- 77.8%
Strengths
- Ranks #2 of 50 compared models
- 98th percentile on this benchmark
- C evidence result
Considerations
- This result measures SWE-Bench Pro, not total model capability
- 03ANClaude Opus 4.8Anthropic
- Score
- 69.2%
- Price
- $5.0 input / $25 output per 1M tokens
- Speed
- Up to 46 tok/s via Anthropic
Strengths
- Ranks #3 of 50 compared models
- 96th percentile on this benchmark
- C evidence result
Considerations
- This result measures SWE-Bench Pro, not total model capability
- 04ACQwen3.8 MaxAlibaba Cloud / Qwen Team
- Score
- 67.7%
- Price
- $2.0 input / $6.0 output per 1M tokens
Strengths
- Ranks #4 of 50 compared models
- 94th percentile on this benchmark
- C evidence result
Considerations
- This result measures SWE-Bench Pro, not total model capability
- 05XAGrok 4.5xAI
- Score
- 64.7%
- Price
- $2.0 input / $6.0 output per 1M tokens
- Speed
- Up to 4.3 tok/s via xAI
Strengths
- Ranks #5 of 50 compared models
- 92th percentile on this benchmark
- C evidence result
Considerations
- This result measures SWE-Bench Pro, not total model capability
Selection summary
Best AI Models for SWE-Bench Pro
Claude Fable 5 currently leads SWE-Bench Pro with 80.0%. 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 SWE-Bench Pro?
What SWE-Bench Pro measures and how its scores work.
SWE-Bench Pro is an advanced version of SWE-Bench that evaluates language models on complex, real-world software engineering tasks requiring extended reasoning and multi-step problem solving.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- SWE-Bench Pro
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- swe-bench-pro|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about SWE-Bench Pro.
Which model scores highest on SWE-Bench Pro?
Claude Fable 5 is currently ranked first with 80.0%.
What does SWE-Bench Pro measure?
SWE-Bench Pro is an advanced version of SWE-Bench that evaluates language models on complex, real-world software engineering tasks requiring extended reasoning and multi-step problem solving.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
50 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.