Benchmark
ASI-Bench
What is ASI-Bench?
First benchmark that jointly evaluates general intelligence, innovation, and autonomous execution, via 60 project-level scientific research tasks spanning 11…
- Released
- 2026-08-18
- Evaluates
- science
- Openness
- unknown
- Importer
- Model reports
- Reported scores
- 18
Source provenance
- Original evidence https://arxiv.org/abs/2608.17271
ASI-Bench paper
ASI-Bench results and reported scores
Model reports
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Claude Opus 4.8 | Anthropic | 31.73 | 2026-08-18 | evidence |
| Claude Opus 5 | Anthropic | 40.7 | 2026-08-18 | evidence |
| DeepSeek-V4-Flash | DeepSeek | 24.84 | 2026-08-18 | evidence |
| DeepSeek-V4-Flash | DeepSeek | 21.91 | 2026-08-18 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 19.2 | 2026-08-18 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 15.78 | 2026-08-18 | evidence |
| GLM-5.2 | Zhipu AI | 30.29 | 2026-08-18 | evidence |
| GLM-5.3 | Zhipu AI | 33.09 | 2026-08-18 | evidence |
| GPT-5.5 (xhigh) | OpenAI | 29.29 | 2026-08-18 | evidence |
| GPT-5.6 Sol (ultra) | OpenAI | 51.6 | 2026-08-18 | evidence |
| GPT-5.6 Sol (xhigh) | OpenAI | 40.86 | 2026-08-18 | evidence |
| Kimi K2.7 | Moonshot AI | 19.75 | 2026-08-18 | evidence |
| Kimi K2.7 | Moonshot AI | 15.65 | 2026-08-18 | evidence |
| Kimi K3 | Moonshot AI | 27.78 | 2026-08-18 | evidence |
| Kimi K3 | Moonshot AI | 28.15 | 2026-08-18 | evidence |
| MiMo-V2.5-Pro | Xiaomi | 16.49 | 2026-08-18 | evidence |
| MiMo-V2.5-Pro | Xiaomi | 11.5 | 2026-08-18 | evidence |
| MiniMax M3 | MiniMax | 20.61 | 2026-08-18 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.
Which sources cite ASI-Bench?
- ASI-Bench results table Apexin AI