← Founder Notes
Archive

The instruction follower just switched flags. minimax m3 leads ifbench at 82.9 percent, edging grok…

Yethikrishna ROriginal on Threads

the instruction follower just switched flags. minimax m3 leads ifbench at 82.9 percent, edging grok 4.3's 81.3 percent across 117 scored models as of oct 9.

the best-behaved model is no longer american.

Context

BenchLeader's IFBench page, as of October 9, lists Grok 4.3 (medium) first at 83.3 percent, Grok 4.20 (thinking) second at 82.9 percent, and MiniMax M3 third at 82.9 percent, with Grok 4.3 (high) at 81.3 percent, across 415 model configurations.

Third-party reviews of MiniMax M3 also give 82.9 percent on IFBench.

How it compares

MiniMax M3 at 82.9 percent and Grok 4.3 at 81.3 percent match the leaderboard figures. The page lists Grok 4.3 (medium) at 83.3 percent first, so 'leads' is not supported there, and the model count is 415, not 117.

Whether another leaderboard shows M3 ahead was not seen, so that is unsupported here, not refuted.

'The best-behaved model is no longer American' is the author's opinion.

Related work

Watch next

  • Read the leaderboard's configuration notes.

Sources

  1. BenchLeader, Oct 9, 2026benchleader.com
  2. OrcaRouter, Sep 23, 2026orcarouter.ai
  3. The Expert Ranking, Sep 23, 2026theexpertranking.com

Provenance

The note above is reproduced unedited from the original post, first published on Threads on 10 October 2026 at 09:58 IST. Sources are the papers and datasets the note draws on.

View the original post
Embed this note
<iframe src="https://founder.myndlabs.tech/notes/embed/the-instruction-follower-just-switched-flags-minimax-m3-DeTOcYvCHfC" width="480" height="420" style="border:0;max-width:100%" loading="lazy" title="The instruction follower just switched flags. minimax m3 leads ifbench at 82.9 percent, edging grok…"></iframe>

More notes