GLM-5.3 vs DeepSeek V4.1 Flash (max)
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. GLM-5.3 leads overall, 88.3 to 88.0. Their confidence intervals overlap, so treat the order as close rather than settled.
The verdict
This one is closer than the ranks suggest. GLM-5.3 sits at #7 with 2256.0 Elo and 88.34 overall; DeepSeek V4.1 Flash (max) is at #14 with 2226.8 Elo and 88.03 overall. Their 95% Elo intervals overlap (2221.7 to 2291.5 against 2185.5 to 2260.8), so do not read much into the exact order. Where GLM-5.3 pulls ahead: Hook Strength (89.95 vs 89.07) and Substance, Accuracy & Value (87.94 vs 87.15). DeepSeek V4.1 Flash (max) still wins on Length Adherence (91.98 vs 90.43) and Visual Cue Quality (83.68 vs 83.02), so it is not a clean sweep. The trade is price: DeepSeek V4.1 Flash (max) costs about $0.013 per article against $0.114 for GLM-5.3. Both publish open weights. Either is a reasonable default: lean GLM-5.3 for hook strength and substance, DeepSeek V4.1 Flash (max) for length adherence, visual cues, and the lower price.
Pick DeepSeek V4.1 Flash (max) for length adherence, visual cues, and the lower price.
Pick GLM-5.3 for the stronger board result, hook strength, and substance.
Metric by metric
Blue bars: GLM-5.3. Orange bars: DeepSeek V4.1 Flash (max). Same 0–100 scale; the bold bar wins that metric.
Everything else that differs
| GLM-5.3 | DeepSeek V4.1 Flash (max) | |
|---|---|---|
| Overall / 100 | 88.3 | 88.0 |
| Writing Elo | 2256 | 2227 |
| Run-to-run spread (± overall std) | 1.480 | 1.280 |
| Cost per script (USD) | 0.114 | 0.013 |
| Avg latency (s) | 312.8 | 76.2 |
| Open weights | Yes | Yes |
Full scorecards: GLM-5.3 · DeepSeek V4.1 Flash (max). How scoring works: methodology.
← All comparisons