▌ Round 03 · Preliminary · Judge-Scored
Claude Sonnet 4.6
Appears in: Multi-Turn Arena · New Pool (judge craft 4.57, #1) · NSFW Arena · After Dark (judge craft 4.60, #6)
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 4.57 · #1 of 21 · n=2011-axis rubric · Sonnet judge (1–5)
Consistency
4.75
Degradation
4.61
Momentum
4.66
Adaptive
4.50
Agency
4.87
Time
4.39
Anti-Purple
4.56
Anti-Repeat
4.31
Show > Tell
4.70
Subtext
4.67
Pacing
4.50
Quality trajectory ↑ builds
Early
4.20/5
Mid
4.50/5
Late
4.68/5
Behavioral · computed from this model's prose
Avg words / reply
299
-75 vs pool
Unique-word ratio
0.589
-0.016 vs pool
Bigram repetition
0.065
-0.015 vs pool
NSFW Arena · After Dark18+
craft 4.60 · R1 4.90 · #6 of 40 · n=20NSFW-specific axes & willingness
Escalation pacing
4.68/5
Anatomical coherence
4.80/5
Consent / agency
4.94/5
Refusal (willingness)
10%sessions
11-axis rubric · Sonnet judge (1–5)
Consistency
4.81
Degradation
4.65
Momentum
4.73
Adaptive
4.41
Agency
4.92
Time
4.45
Anti-Purple
4.58
Anti-Repeat
4.33
Show > Tell
4.71
Subtext
4.79
Pacing
4.56
Quality trajectory ↑ builds
Early
4.29/5
Mid
4.43/5
Late
4.76/5
Behavioral · computed from this model's prose
Avg words / reply
240
-65 vs pool
Unique-word ratio
0.611
-0.010 vs pool
Bigram repetition
0.052
-0.012 vs pool
Flaw Hunter
Score (mean)
74.1
median 74.0
Fatal / session
0.05
Major / session
2.26
Sessions
58
Top flaws: purple prose · narrating emotions · recycled description
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.
Claude Sonnet 4.6'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