Coding benchmark
Terminal-Bench 2.1 leaderboard
Terminal-Bench 2.1 (provider run). Every model the catalog carries a published Terminal-Bench 2.1 value for, ranked by that value.
CategoryCoding
MeasureInteractive task success rate
TasksTerminal-based software-agent tasks
DifficultyProfessional software engineering
Published byDeepSeek V4 Flash 0731 update
A provider-run Terminal-Bench 2.1 result stored separately from the repository's Terminal-Bench 2.0 lane.
Terminal-Bench 2.1 ranking
33 models with a published Terminal-Bench 2.1 value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Interactive task success rate |
|---|---|---|---|
| 1 |  SWE-2 | Cognition | 92.8 |
| 2 |  DeepSeek V4.1 Flash | DeepSeek | 90.6 |
| 3 |  Gemini 3.8 Flash | 89.4 | |
| 4 |  Muse Spark 1.3 | Meta | 88.8 |
| 5 |  GLM-5.3 | Z.AI | 88.2 |
| 6 | DeepSeek V4 Pro 0813 | DeepSeek | 87.9 |
| 7 |  Qwen3.8 Max | Alibaba | 86.6 |
| 8 | OAOrnith-1.5-397B | Ornith AI | 86.1 |
| 9 | Gemini 3.7 Flash | 85.8 | |
| 10 |  Hy4 preview | Tencent | 85.4 |
| 11 | GLM-5.3-Flash | Z.AI | 84.3 |
| 12 | Muse Spark 1.2 | Meta | 82.9 |
| 13 | DeepSeek V4 Flash 0731 | DeepSeek | 82.7 |
| 14 |  Atria Dawn Preview | Shanghai Artificial Intelligence Laboratory | 78.3 |
| 15 | DSdots3-note Preview | Dots Studio | 75.1 |
| 16 | Qwen3.8-27B | Alibaba | 73.0 |
| 17 |  Seed 2.1 Pro | ByteDance | 71.0 |
| 18 | APApodex 1.1 | Apodex | 70.8 |
| 19 | MCQuasar 438B | Multiverse Computing | 69.3 |
| 20 | OAOrnith-1.5-35B-A3B | Ornith AI | 67.8 |
| 21 | Seed 2.1 Turbo | ByteDance | 67.6 |
| 22 | PAPokee-Isaac 28B | Pokee AI | 65.1 |
| 23 |  MAI-Code-1.1-Flash | Microsoft | 62.9 |
| 24 |  Ling 3.0 Flash | InclusionAI | 57.0 |
| 24 |  Solar Pro 4 | Upstage | 57.0 |
| 26 | PMTernary Bonsai 2 27B | Prism ML | 52.8 |
| 27 | Muse Glimmer 30B | Meta | 51.7 |
| 28 | OAOrnith-1.5-9B | Ornith AI | 46.2 |
| 29 |  K-EXAONE 2.0 | LG AI Research | 43.8 |
| 30 | STA.X K2 | SK Telecom | 36.0 |
| 31 |  Granite 4.2 30B | IBM | 29.2 |
| 32 | Granite 4.2 8B | IBM | 20.6 |
| 33 | OPMiniCPM5-2B | OpenBMB | 8.6 |
Evidence key: Observed
Rows are ordered by the value DeepSeek V4 Flash 0731 update published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.