DOOMBENCH.COM
Hourly watch
Autonomy and agency

DeepSeek releases hybrid-reasoning DeepSeek-V3.1

DeepSeek released V3.1 with switchable thinking and non-thinking modes, stronger tool use and agent capabilities, an API and public model weights.

CURRENT ASSESSMENT · REVISION 1
TOWARD DOOM54confidence 92/100

Why it moved the index

Public weights and improved tool use for an exact hybrid-reasoning checkpoint materially expanded accessible agentic capability and diffusion.

0 comments · 0 votesOpen discussion

Public discussion is readable by everyone. Sign in to comment, reply, or vote.

No comments yet. Start the discussion.

Send detailed or private feedback →
AUDIT TRAIL

Assessment history

  1. R1
    Toward 54 · confidence 92

    New exact August 2025 model release with authoritative repository timestamp and no durable collision.

    12 Aug 2026