▌ Round 03 · Preliminary · Judge-Scored
DeepSeek V3 0324
Appears in: Multi-Turn Arena · New Pool (judge craft 3.90, #13) · NSFW Arena · After Dark (judge craft 4.37, #29)
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 3.90 · #13 of 21 · n=2011-axis rubric · Sonnet judge (1–5)
Consistency
3.93
Degradation
3.58
Momentum
4.16
Adaptive
3.90
Agency
4.40
Time
3.98
Anti-Purple
4.04
Anti-Repeat
3.36
Show > Tell
4.38
Subtext
4.16
Pacing
3.92
Quality trajectory ↓ degrades
Early
4.05/5
Mid
3.92/5
Late
3.88/5
Behavioral · computed from this model's prose
Avg words / reply
207
-167 vs pool
Unique-word ratio
0.720
+0.115 vs pool
Bigram repetition
0.038
-0.042 vs pool
NSFW Arena · After Dark18+
craft 4.37 · R1 4.90 · #29 of 40 · n=20NSFW-specific axes & willingness
Escalation pacing
4.75/5
Anatomical coherence
4.64/5
Consent / agency
4.92/5
Refusal (willingness)
0%sessions
11-axis rubric · Sonnet judge (1–5)
Consistency
4.54
Degradation
4.17
Momentum
4.43
Adaptive
4.47
Agency
4.78
Time
4.21
Anti-Purple
4.18
Anti-Repeat
3.95
Show > Tell
4.50
Subtext
4.41
Pacing
4.35
Quality trajectory ↑ builds
Early
4.20/5
Mid
4.30/5
Late
4.49/5
Behavioral · computed from this model's prose
Avg words / reply
150
-154 vs pool
Unique-word ratio
0.738
+0.117 vs pool
Bigram repetition
0.022
-0.042 vs pool
Flaw Hunter
Score (mean)
68.1
median 69.0
Fatal / session
0.10
Major / session
2.71
Sessions
58
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.
DeepSeek V3 0324'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