DeepSeek V4 Flash 0731 achieved 89.0% accuracy on the ARC-AGI-1 semi‑private benchmark at a cost of $0.02 per task and 61.4% on the ARC-AGI-2 semi‑private benchmark at $0.04 per...
DeepSeek V4 Flash 0731 achieved 89.0% accuracy on the ARC-AGI-1 semi‑private benchmark at a cost of $0.02 per task and 61.4% on the ARC-AGI-2 semi‑private benchmark at $0.04 per task, according to the ARC Prize leaderboard. The model, released July 31, 2026, is part of DeepSeek’s V4 series. The leaderboard lists three performance tiers—Max, High, and Low—showing the Max variant scoring 89.0% on ARC‑AGI‑1 and 61.4% on ARC‑AGI‑2, while the High and Low variants scored lower, with the Low variant reaching 84.0% on ARC‑AGI‑1 and 46.0% on ARC‑AGI‑2. The public eval comprises 120 ARC‑AGI‑2 tasks and 400 ARC‑AGI‑1 tasks, each categorized by difficulty. The results place DeepSeek V4 Flash among the top‑performing models in the current competition.
- Publisher
- Hacker News
- Reliability
- high
- Published
- 8/8/2026, 10:05:11 AM
- Retrieved
- 8/8/2026, 10:05:11 AM
- Relevance
- 80%
- Confidence
- 85%

