▌ Round 03 · Preliminary · Judge-Scored
Qwen 3.6 27B
Appears in: Multi-Turn Arena · New Pool (judge craft 4.25, #11) · NSFW Arena · After Dark (judge craft 4.45, #23)
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 4.25 · #11 of 21 · n=1911-axis rubric · Sonnet judge (1–5)
Consistency
4.51
Degradation
4.11
Momentum
4.33
Adaptive
4.29
Agency
4.75
Time
4.26
Anti-Purple
4.17
Anti-Repeat
3.74
Show > Tell
4.45
Subtext
4.28
Pacing
4.16
Quality trajectory → holds
Early
4.16/5
Mid
4.24/5
Late
4.24/5
Behavioral · computed from this model's prose
Avg words / reply
284
-90 vs pool
Unique-word ratio
0.653
+0.048 vs pool
Bigram repetition
0.049
-0.031 vs pool
NSFW Arena · After Dark18+
craft 4.45 · R1 4.96 · #23 of 40 · n=20NSFW-specific axes & willingness
Escalation pacing
4.81/5
Anatomical coherence
4.72/5
Consent / agency
4.95/5
Refusal (willingness)
0%sessions
11-axis rubric · Sonnet judge (1–5)
Consistency
4.68
Degradation
4.30
Momentum
4.57
Adaptive
4.64
Agency
4.87
Time
4.28
Anti-Purple
4.26
Anti-Repeat
3.89
Show > Tell
4.47
Subtext
4.52
Pacing
4.42
Quality trajectory ↑ builds
Early
4.28/5
Mid
4.45/5
Late
4.42/5
Behavioral · computed from this model's prose
Avg words / reply
318
+14 vs pool
Unique-word ratio
0.618
-0.003 vs pool
Bigram repetition
0.057
-0.007 vs pool
Flaw Hunter
Score (mean)
66.4
median 69.0
Fatal / session
0.04
Major / session
3.22
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 absolute flaw scores across new vs returning models with that caveat in mind.
This is the judges' call — not the crowd's.
Qwen 3.6 27B'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