AI BriefingDeepSeek V4 Flash 0731 ARC-AGI Evaluation Results
·2026.08.08 06:36
Key point
DeepSeek V4 Flash 0731 achieved 89.0% on ARC-AGI-1.
Details
The ARC-AGI evaluation results for V4 Flash 0731, released by DeepSeek, have been published.
The following performance was recorded under the maximum inference setting.
- ARC-AGI-1 Semi-Private: 89.0%, $0.02 per task
- ARC-AGI-2 Semi-Private: 61.4%, $0.04 per task
Scores by inference level are as follows.
- Max: ARC-AGI-1 89.0%, ARC-AGI-2 61.4%
- High: ARC-AGI-1 87.0%, ARC-AGI-2 56.0%
- Low: ARC-AGI-1 84.0%, ARC-AGI-2 46.0%
The evaluation page also presents pass/fail results by inference level for the ARC-AGI-1 public evaluation 400 tasks and the ARC-AGI-2 public evaluation 120 tasks. The ARC-AGI-3 score was not disclosed.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
AI Agent CAD Benchmark: CadQuery's Explicit Failure vs OpenSCAD's Silent FailureThe Pitfalls of OpenRouter Routing: Performance and Feature Disparities Across Providers for the Same Model