The API now answers MiniMax-M3.1-Flash-Preview as MiniMax-M3.1-Flash, and
accepts the new name directly. Verified 2026-11-03: a live call returns
"ok" with a 1M context window. Claude Code 2.1.317 prints an
unrecognized_model warning for either name; the session still runs.
Claude-Session: https://claude.ai/code/session_01CLUCNYsiEaER1CMqw5WqfY
Flash is the default model; --m3 falls back to M3, --effort overrides.
Defaults: max (review-minimax, review-loop, mm-worktree-task), medium
(explore), low (ci-triage). ask-minimax-run applies MINIMAX_EFFORT to
Flash models when --effort is not passed. Bump stale fixture date.