DeepSeek 4.1 Flash outputs 88K tokens in a single task

Charts from Artificial Analysis show DeepSeek 4.1 Flash output 88,574 tokens in one benchmark task; users also found its thinking effort is tunable from 1 to 100, defaulting to 50.

2026-09-11 ~ 2026-09-11 · 2 related posts

• DeepSeek 4.1 Flash emits 88,574 tokens per task vs 4,432 for GPT-6 Astra low, benchmark shows — Tight-Grocery9053 · 2026-09-11
• DeepSeek V4.1's thinking effort is tunable 1-100; 'low effort' is just 50, tester finds — teortaxesTex · 2026-09-11