All 1 scoreable measurements here come from one runner under one declared configuration, so nothing on this page explains why they differ. That is the finding.
Published by the model's own maker, on a variant of this benchmark that no independent party we hold has run. They are recorded and they are shown. They can never move a rank.
| Variant | Claimed |
|---|---|
| 100_steps_one_run | 75.2 |
| 361_samples_200_steps | 70.1 |
3 rows. Best first. the board takes the LOWER MEDIAN of the scoreable measurements, which is a real row with its own interval and provenance rather than an interpolated average. a vendor self-report is shown beside the independent numbers and is never one of them; it cannot move a rank.
| Score | Provenance | Configuration | Who ran it | Where to check it |
|---|---|---|---|---|
| 75.2 / 100
as published: 75.2% |
CLAIMED
shown, never scored |
MiniMax-M3 harness MiniMax-submitted official leaderboard row effort not_disclosed variant 100_steps_one_run |
MiniMax via Model benchmark dossier 2026-06-08T00:00:00.000Z facts, not expression |
checked: this figure appears in the page’s row for this model2026-08-30 yutori.comMiniMax M3 | MiniMax | General model | 100 | 75.2% | 0.0 | 1 | Jun 08, 2026 | No | No | No |
| 75.2 / 100
as published: 75.2% |
INDEPENDENT
counts toward the ranking |
MiniMax M3 harness General model effort 100 steps |
OSWorld (XLANG Lab) via OSWorld leaderboard none stated |
NO_URL2026-08-30 benchmark leaderboardthe leaderboard for this benchmark, not a link to this row |
| 70.1 / 100
as published: 70.1% |
CLAIMED
shown, never scored |
MiniMax-M3 harness official OSWorld-Verified code effort not_disclosed variant 361_samples_200_steps |
MiniMax via Model benchmark dossier 2026-06-01T00:00:00.000Z facts, not expression |
checked: this figure appears in the page’s row for this model2026-08-30 minimax.ioVideoMMMU: Video frame rate = 1 FPS, max 512 frames, single-frame long edge 672–1008 pixels. Official VideoMMMU prompt used, LLM-as-a-Judge scoring. MiniMax M3: |
Snapshot 2026-08-30-62e4e85f043b, manifest dc46131315046f1e, as of 2026-08-30. where only one party has measured a model on a benchmark, this page says so rather than presenting a single number as a reconciliation.