▌ Round 03 · Preliminary · Judge-Scored
Magnum v4 72BRP finetune
Appears in: Multi-Turn Arena · New Pool (judge craft 3.41, #17) · NSFW Arena · After Dark (judge craft 3.23, #36)
Judge-scored only — no human arena votes yet. These numbers may move (or invert) once the human arena fills in. Pre-vote, there is no human ELO, composite, or cost/latency for this model.
Multi-Turn Arena · New Pool
craft 3.41 · #17 of 21 · n=2011-axis rubric · Sonnet judge (1–5)
Consistency
3.58
Degradation
2.92
Momentum
3.62
Adaptive
3.76
Agency
4.12
Time
3.85
Anti-Purple
3.29
Anti-Repeat
2.61
Show > Tell
3.50
Subtext
3.69
Pacing
3.40
Quality trajectory ↓ degrades
Early
4.03/5
Mid
3.64/5
Late
3.10/5
Behavioral · computed from this model's prose
Avg words / reply
316
-57 vs pool
Unique-word ratio
0.596
-0.009 vs pool
Bigram repetition
0.081
+0.001 vs pool
NSFW Arena · After Dark18+
craft 3.23 · R1 3.98 · #36 of 40 · n=19NSFW-specific axes & willingness
Escalation pacing
3.83/5
Anatomical coherence
4.00/5
Consent / agency
4.28/5
Refusal (willingness)
0%sessions
11-axis rubric · Sonnet judge (1–5)
Consistency
3.33
Degradation
2.64
Momentum
3.50
Adaptive
3.56
Agency
4.18
Time
3.81
Anti-Purple
2.96
Anti-Repeat
2.24
Show > Tell
3.30
Subtext
3.47
Pacing
3.18
Quality trajectory ↓ degrades
Early
3.97/5
Mid
3.31/5
Late
2.82/5
Behavioral · computed from this model's prose
Avg words / reply
255
-49 vs pool
Unique-word ratio
0.601
-0.020 vs pool
Bigram repetition
0.080
+0.016 vs pool
Flaw Hunter
Score (mean)
52.1
median 55.0
Fatal / session
0.52
Major / session
3.83
Sessions
58
Top flaws: recycled description · purple prose · narrating emotions
From the Round 03 flaw-hunter catch-up run. Scores come from two runs — compare absolute flaw scores across new vs returning models with that caveat in mind.
This is the judges' call — not the crowd's.
Magnum v4 72B's scores here are judge-only. Read a session it played and vote — the human ranking takes shape on the leaderboard as ballots land.
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