deepseek-v4.1IMAGE
ARC Prize benchmarks DeepSeek V4.1 Flash from low to max reasoning
ARC-AGI-1 rose from 97.2% to 98.4% with reasoning tokens growing from 26k to 34k per task, while ARC-AGI-2 improved from 75.6% to 79.2% with tokens rising from 108k to 134k.
ARC Prize
@arcprizeArchived 9 days agoOriginally posted 2026.10.23
ARC Prize @arcprizeOriginally posted 2026.10.23
@deepseek_ai DeepSeek V4.1 Flash 1017 saw modest gains from low to max on public tasks: ARC-AGI-1 rose from 97.2% to 98.4%, while reasoning grew from 26k to 34k tokens/task. On ARC-AGI-2, the gain was larger: 75.6% to 79.2%, with reasoning growing from 108k to 134k tokens/task.
x.com/arcprize/status/2113318035054321673Archived 2026.10.23 12:08