Screenshot surfaces 'DeepSeek-V4.1-Flash Max' with 99.7% cache hit rate
mariofilhoml · x · 2026-11-10
A screenshot from a dev tool shows a 'DeepSeek-V4.1-Flash Max' model option, with 37.1M tokens processed, a 99.7% cache hit rate, and 29% progress. Possibly an unannounced new DeepSeek model or version surfaced early — third-party screenshot, unverified.
More from Models
- Gemini silently downgrades Pro users to Flash-Lite after hitting limits — CleitonFenix · 2026-11-10
- LLM tokenizers repeat the 1870s Chinese telegraph problem: up to 17x cost gap across languages — demian_ai · 2026-11-10
- MiniMax H3 Image open-weights image model praised for speed, quality and novel aesthetics — multimodalart · 2026-11-10
- Three Claude models draw the human body in motion from the same prompt — claudeai · 2026-11-10
- Anthropic shows early Claude Fable 5.2 experiments: a 3D cell model built entirely from code — claudeai · 2026-11-10
- If a model scores below Kimi K3, blame the benchmark, not the model — nitarshan · 2026-11-10