DeepSeek V4.1-Flash is live on Nito.
Run it inside Claude Code or Codex, or in the Nito web app, fully incognito. By DeepSeek's numbers it tops Opus 5 and GPT-5.6 on all four agentic benchmarks, with 8B active parameters on input and a KV cache a quarter the size.
DeepSeek Flash 4.1 on my NYC test.
V4 non flash in the middle, Opus 5 underneath. In April the full V4 gave me a black screen. Thats the middle panel, same prompt. Flash 4.1 is a massive improvement for the size. Golden hour, water with reflections, lit windows, water towers on
The lead anthropic engineer leaked a system where one brain runs 40 agents and builds a business doing $100,000 a month
One Anthropic brain instead of forty opinions - it costs nothing and produces $1.2M a year brief → plan → shift → review → rewrite → back to the brief
Time to let you go AnthropicAI
Grok 4.6 high outperformed your Opus 5 xhigh with better and faster results.
You have reached the end of the archive
All of Claude Opus 5