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. Grok 4.5 leads overall, 85.8 to 82.4.
A closer fight than the ranks suggest, and the first thing to know is that cost disappears from this decision: both land around two cents per script, GLM's estimated. Grok 4.5 wins, and even with only two points between them overall, the Elo intervals don't overlap, so I'll trust the order. Where they actually split is numbers. Grok scores 85.6 on our anti-slop and numbers check while GLM-5 sits at 77.3, meaning GLM is the one reciting stat sheets at you. GLM punches back on length adherence, 87.1 to Grok's 78.7, and Grok's variance there is high, so expect the occasional draft that overshoots badly. On tone the gap is a couple of points with standard deviations to match, so honestly, inside the noise. Two practical tiebreakers: Grok returns a script in under 30 seconds, the fastest here, while GLM is open weights, which Grok can't offer. Both trail my reference scripts by a wide margin, to be clear. For quality per dollar, Grok, narrowly.
Pick Grok 4.5 if you want cleaner number handling and near-instant drafts at the same two-cent price.
Pick GLM-5 if open weights matter and you'd rather fix the occasional stat dump than fix length overruns.
Blue bars: Grok 4.5. Orange bars: GLM-5. Same 0–100 scale; the bold bar wins that metric.
| Grok 4.5 | GLM-5 | |
|---|---|---|
| Overall / 100 | 85.8 | 82.4 |
| Writing Elo | 2042 | 1783 |
| Run-to-run spread (± overall std) | 2.810 | 3.760 |
| Cost per script (USD) | 0.038 | 0.023 |
| Avg latency (s) | 41.7 | 117.3 |
| Open weights | No | Yes |
Full scorecards: Grok 4.5 · GLM-5. How scoring works: methodology.
← All comparisons