▌ Round 03 · Preliminary · Judge-Scored
MiniMax M3
Appears in: Multi-Turn Arena · New Pool (judge craft 4.46, #3) · NSFW Arena · After Dark (judge craft 4.58, #9)
Judge-scored only: no human arena votes are included here. The human ranking can move or invert these numbers. There is no human ELO, composite, or cost/latency for this model on this page.
Multi-Turn Arena · New Pool
craft 4.46 · #3 of 21 · n=2011-axis rubric · Sonnet judge (1–5)
Consistency
4.71
Degradation
4.44
Momentum
4.61
Adaptive
4.45
Agency
4.80
Time
4.33
Anti-Purple
4.34
Anti-Repeat
4.15
Show > Tell
4.63
Subtext
4.55
Pacing
4.41
Quality trajectory ↑ builds
Early
4.17/5
Mid
4.48/5
Late
4.58/5
Behavioral · computed from this model's prose
Avg words / reply
512
+138 vs pool
Unique-word ratio
0.535
-0.070 vs pool
Bigram repetition
0.097
+0.017 vs pool
NSFW Arena · After Dark18+
craft 4.58 · R1 4.89 · #9 of 40 · n=20NSFW-specific axes & willingness
Escalation pacing
4.72/5
Anatomical coherence
4.76/5
Consent / agency
4.84/5
Refusal (willingness)
5%sessions
11-axis rubric · Sonnet judge (1–5)
Consistency
4.81
Degradation
4.56
Momentum
4.72
Adaptive
4.55
Agency
4.92
Time
4.44
Anti-Purple
4.45
Anti-Repeat
4.21
Show > Tell
4.69
Subtext
4.77
Pacing
4.54
Quality trajectory ↑ builds
Early
4.23/5
Mid
4.46/5
Late
4.72/5
Behavioral · computed from this model's prose
Avg words / reply
555
+251 vs pool
Unique-word ratio
0.500
-0.121 vs pool
Bigram repetition
0.116
+0.052 vs pool
Flaw Hunter
Craft band
A band, not a number: about ten points either way is rater noise.
Fatal / session
0.15
Major / session
2.80
Sessions
55
Top flaws: purple prose · recycled description · narrating emotions
From the Round 03 flaw-hunter catch-up run. Scores come from two runs; compare flaw bands across new and returning models with that caveat in mind. The current model cards carry the newer single-rater band.
This is the judges' call, not the crowd's.
MiniMax M3's scores here are judge-only. Round 03's human voting has closed; its board holds the crowd's ranking.
Source: rp-benchmark · scripts/build-round3-judge-preview.py. Rubric & trajectory are Claude Sonnet judge means; behavioral is computed from the model's own dialogue. No human votes included.
← Back to the Round 03 standings