Towards AITowards AIToneBench

Claude Fable 5 (max) vs Gemini 3.1 Pro (default)

Head-to-head on the Towards AI writing benchmark: same 9 real YouTube scripts, five runs each, scored blind by a three-family judge panel. Claude Fable 5 (max) leads overall, 90.3 to 80.2.

Claude Fable 5 (max)
#2 Elo 2575 · 90.3/100
Gemini 3.1 Pro (default)
#63 Elo 1577 · 80.2/100
Cost / script
$0.908 vs $0.118
Human baseline
92.7 Claude Fable 5 (max) falls below it · Gemini 3.1 Pro (default) falls below it

The verdict

I use Gemini daily for image generation and deep-research work, so let me be fair up front: this result is about scriptwriting in my voice, not Gemini overall. And on that job, it's not close. Fable 5 scores 90.21 to Gemini 3.1 Pro's 80.21, the widest gap in this set of matchups, with Gemini sitting at rank 53 on the board. The two tells: writing craft, 80.82 vs 89.6, and length adherence, where Gemini lands at 75.91 with a deviation near 9, meaning scripts regularly come back the wrong size. Continuity and emotion trails too, 78.40 vs 88.95, and that's the metric that decides whether anyone is still watching at minute eight. What Gemini has going for it: speed, about a minute per script, and a mid-tier price around twelve cents. But it costs roughly six times what Grok 4.5 does while ranking well below it, which makes it hard to place. For this specific job, Fable earns the dollar. Gemini, I'll keep for everything else.

Pick Fable 5 if you're publishing the scripts; the craft, pacing, and length gaps here are too wide to edit away cheaply.
Pick Gemini 3.1 Pro if you want fast, one-minute drafts inside a Google stack and treat the output as raw material.

Metric by metric

Blue bars: Claude Fable 5 (max). Orange bars: Gemini 3.1 Pro (default). Same 0–100 scale; the bold bar wins that metric.

Tone & Voice Match19% weight
90.4
79.9
Writing Craft & Clarity13% weight
89.6
80.8
Substance, Accuracy & Value15% weight
90.2
80.8
Continuity & Emotion14% weight
89.1
78.4
YouTube Best Practices12% weight
89.0
78.9
Hook Strength10% weight
90.1
82.3
Length Adherence8% weight
93.5
75.9
Slop Score (EQ-Bench + ours)5% weight
93.7
87.7
Visual Cue Quality4% weight
90.0
81.7

Everything else that differs

Claude Fable 5 (max)Gemini 3.1 Pro (default)
Overall / 10090.380.2
Writing Elo25751577
Run-to-run spread (± overall std)1.6102.420
Cost per script (USD)0.9080.118
Avg latency (s)452.863.8
Open weightsNoNo

Full scorecards: Claude Fable 5 (max) · Gemini 3.1 Pro (default). How scoring works: methodology.

← All comparisons