Capability gains

SpaceXAI unveils Grok 3 and Grok 3 mini reasoning models

SpaceXAI unveiled Grok 3 and Grok 3 mini, beta reasoning models trained with large-scale reinforcement learning and ten times the compute of its prior generation, with user rollout beginning.

CURRENT ASSESSMENT · REVISION 1
TOWARD DOOM67confidence 96/100

Why it moved the index

The release materially advanced reasoning, coding and test-time compute while beginning deployment to users. Grok 3 Think and mini Think are treated as reasoning modes of the two named models, not invented additional immutable checkpoints.

0 comments · 0 votesOpen discussion

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

No comments yet. Start the discussion.

AUDIT TRAIL

Assessment history

  1. R1
    Toward 67 · confidence 96

    New February 2025 release of two exact model tiers with no durable identity collision.

    12 Aug 2026