HyperAI
Command Palette
Search for a command to run...
Terminal-Bench 4.0
소프트웨어, 머신러닝, 과학, 운영, 보안, 하드웨어, 미디어 전반의 복잡한 터미널 작업 66개로 구성된 더 어려운 벤치마크로, 컴퓨트와 시간 허용량을 재조정하고 지침, 환경, 검증기를 개선했습니다.
- 카테고리
- 코딩
- 지능 지수 가중치
- 10%
- 모델
- 147 현재, 252 전체
- 중앙값
- 11.1%
- 범위
- 0.0%–67.2%
| 순위 | 비교 | 모델 | 점수 | 비용 / 작업 | 시간 / 작업 | 출력 토큰 / 작업 |
|---|---|---|---|---|---|---|
| 51 |  Claude Haiku 5.5 (Xhigh) Anthropic | 29.3% | $0.46 | 16m 19s | 232K | |
| 52 |  DeepSeek V4.1 Flash 1017 (High) DeepSeek | 28.8% | $0.78 | 21m 32s | 183K | |
| 53 | 26.8% | $0.38 | 4m 0s | 16.6K | ||
| 54 |  Grok 4.7 (Xhigh) SpaceXAI | 25.8% | $12.43 | 41m 12s | 158K | |
| 55 |  MiniMax-M3.1-Flash MiniMax | 25.8% | $0.92 | 27m 21s | 262K | |
| 56 |  Qwen3.8-Flash-Next Alibaba | 25.3% | — | — | — | |
| 57 | Grok 4.7 (High) SpaceXAI | 24.7% | $7.63 | 31m 13s | 126K | |
| 58 |  Mistral Large 4 (High) Mistral | 24.7% | $4.59 | 24m 20s | 171K | |
| 59 | Mistral Large 4 (Medium) Mistral | 23.2% | $2.89 | 17m 7s | 116K | |
| 60 |  MiMo-V2.6-Flash Xiaomi | 22.7% | $0.23 | 112m 16s | 236K | |
| 61 | Claude Haiku 5.5 (High) Anthropic | 21.7% | $0.33 | 12m 8s | 166K | |
| 62 | 20.7% | $1.28 | 4m 49s | 40.3K | ||
| 63 |  Gemini 3.8 Flash (High) Google | 19.7% | $18.15 | 26m 38s | 184K | |
| 64 | 19.7% | $14.17 | 23m 23s | 157K | ||
| 65 | Grok 4.7 (Low) SpaceXAI | 16.2% | $5.03 | 17m 16s | 66.9K | |
| 66 | Claude Haiku 5.5 (Medium) Anthropic | 15.2% | $0.19 | 7m 53s | 107K | |
| 67 |  GPT-6.1 Terra (Low) OpenAI | 15.2% | $0.11 | 1m 50s | 7.48K | |
| 68 | DeepSeek V4 Pro 0813 (Max) DeepSeek | 14.1% | $3.76 | 41m 14s | 172K | |
| 69 |  Inkling 2 Small Thinking Machines | 13.1% | $2.48 | 37m 53s | 290K | |
| 70 | Claude Haiku 5.5 (Low) Anthropic | 12.6% | $0.12 | 5m 21s | 70.3K | |
| 71 | 12.6% | — | — | — | ||
| 72 | GPT-6 Luna (Max) OpenAI | 12.6% | $0.24 | 21m 30s | 141K | |
| 73 | DeepSeek V4 Flash Vision (Max) DeepSeek | 12.1% | — | — | — | |
| 74 | Qwen3.8 2.4T A95B Alibaba | 11.1% | — | — | — | |
| 75 |  Agnes 2.5 Pro Theta Sapiens AI | 10.1% | $0.79 | 39m 3s | 453K | |
| 76 | Gemini 3.8 Flash (Low) Google | 10.1% | — | — | — | |
| 77 | 9.1% | $0.48 | 13m 33s | 102K | ||
| 78 | GPT-6 Luna (Xhigh) OpenAI | 8.1% | $0.20 | 16m 52s | 104K | |
| 79 | 8.1% | $4.66 | 11m 47s | 122K | ||
| 80 | Qwen3.8 27B (Xhigh) Alibaba | 5.6% | $4.57 | 53m 33s | 148K | |
| 81 | Qwen3.8 27B (Medium) Alibaba | 5.1% | $6.53 | 57m 42s | 155K | |
| 82 | GPT-6 Luna (High) OpenAI | 4.5% | $0.12 | 12m 46s | 74.2K | |
| 83 | Gemini 3.1 Pro Preview Google | 4.0% | — | — | — | |
| 84 |  JT-4.1 Flash 236B A21B (Reasoning) China Mobile | 3.0% | — | — | — | |
| 85 |  Beam 1.0 Reflection | 3.0% | $0.63 | 4m 1s | 69.4K | |
| 86 | GPT-6 Luna (Medium) OpenAI | 2.5% | $0.072 | 7m 32s | 45K | |
| 87 | Qwen3.8 27B (Low) Alibaba | 2.5% | $6.90 | 59m 49s | 180K | |
| 88 | MiniMax-M3 MiniMax | 2.0% | $3.04 | 23m 43s | 177K | |
| 89 | 1.5% | $2.58 | 53m 16s | 212K | ||
| 90 | 1.5% | $0.045 | 2m 32s | 13.7K | ||
| 91 |  K2 Horizon 375B A23B Institute of Foundation Models | 1.5% | $0 | 15m 51s | 179K | |
| 92 | Gemini 3.5 Flash-Lite Google | 1.0% | $1.10 | 2m 56s | 54.8K | |
| 93 | Inkling (Xhigh) Thinking Machines | 1.0% | $2.58 | 15m 0s | 87.7K | |
| 94 | Inkling Small Thinking Machines | 1.0% | $0.35 | 13m 9s | 56.5K | |
| 95 | K2 Horizon 7B Institute of Foundation Models | 1.0% | — | — | — | |
| 96 |  Kimi K2.7 Code Kimi | 1.0% | — | — | — | |
| 97 |  Quasar 438B (Max, Based on GLM-5.2) Multiverse Computing | 1.0% | — | — | — | |
| 98 | Qwen3.7 Plus Alibaba | 1.0% | — | — | — | |
| 99 |  Solar Mini 4 Upstage | 1.0% | — | — | — | |
| 100 |  Command A+ Cohere | 0.5% | $0 | 4m 11s | 50K |
- 52