GPT-5.6 Sol (high) vs Kimi K3 (thinking)

Head-to-head on the Towards AI writing benchmark: same 10 real YouTube scripts, five runs each, scored blind by a three-family judge panel. GPT-5.6 Sol (high) leads overall, 87.6 to 87.2. Their per-model Elo intervals overlap on this board.

GPT-5.6 Sol (high)

#33 Elo 2134 · 87.6/100

Kimi K3 (thinking)

#41 Elo 2098 · 87.2/100

Cost / script

$0.155 vs $0.095

Human baseline

90.9 GPT-5.6 Sol (high) falls below it · Kimi K3 (thinking) falls below it

The verdict

This one is closer than the ranks suggest. GPT-5.6 Sol (high) sits at #33 with 2133.6 Elo and 87.61 overall; Kimi K3 (thinking) is at #41 with 2098.0 Elo and 87.21 overall. Their 95% Elo intervals overlap (2093.6 to 2169.6 against 2058.5 to 2130.1), so do not read much into the exact order. Where GPT-5.6 Sol (high) pulls ahead: Visual Cue Quality (89.38 vs 83.52) and Length Adherence (89.30 vs 86.27). Kimi K3 (thinking) still wins on Hook Strength (88.27 vs 86.56) and Tone & Voice Match (87.32 vs 86.52), so it is not a clean sweep. The trade is price: Kimi K3 (thinking) costs about $0.095 per article against $0.155 for GPT-5.6 Sol (high), so you pay for the higher rank. Kimi K3 (thinking) publishes open weights; GPT-5.6 Sol (high) does not. Default to GPT-5.6 Sol (high). Reach for Kimi K3 (thinking) when the hook, tone, the lower price, and open weights matter more to you than the headline rank.

Pick GPT-5.6 Sol (high) for the stronger board result, visual cues, and length adherence.
Pick Kimi K3 (thinking) for the hook, tone, the lower price, and open weights.

Reading the Elo intervals. Each range estimates uncertainty around one model's Elo. We repeatedly sample from every model's recorded run scores on each task and recalculate the full ranking. Comparing these separate ranges does not test the Elo gap between two models. The tasks stay fixed, so the ranges do not measure how the ranking would change on new tasks.

Metric by metric

Blue bars: GPT-5.6 Sol (high). Orange bars: Kimi K3 (thinking). Same 0–100 scale; the bold bar wins that metric.

Tone & Voice Match19% weight

86.5

87.3

Writing Craft & Clarity13% weight

88.1

88.3

Substance, Accuracy & Value15% weight

88.2

87.1

Continuity & Emotion14% weight

86.1

86.4

YouTube Best Practices12% weight

86.8

Hook Strength10% weight

86.6

88.3

Length Adherence8% weight

89.3

86.3

Slop Score (EQ-Bench + ours)5% weight

92.8

89.8

Visual Cue Quality4% weight

89.4

83.5

Everything else that differs

	GPT-5.6 Sol (high)	Kimi K3 (thinking)
Overall / 100	87.6	87.2
Writing Elo	2134	2098
Run-to-run spread (± overall std)	1.330	1.550
Cost per script (USD)	0.155	0.095
Avg latency (s)	82.7	70.9
Open weights	No	Yes

Full scorecards: GPT-5.6 Sol (high) · Kimi K3 (thinking). How scoring works: methodology.

← All comparisons