Grok 3
xAI · proprietary · writing benchmark
The short version
Grok 3 sits at #103 of 130 on ToneBench, toward the value end of the board, with a writing Elo of 1016 and an overall score of 70.5 out of 100. We measured it by having it write all 9 of our real YouTube scripts, five times each, then scoring every draft blind against our own finished versions. Here is how it shook out.
Within xAI it ranks 6th of 7. Grok 4.5 is the family's top writer here, about 1026 Elo ahead, so this config trades some quality for whatever it saves you in cost or speed.
Among proprietary models, it comes in 76th of 83. That ranking is against the stronger half of the board: closed models still set the pace on voice fidelity here.
Worth knowing before you rely on it: this model is swingy. Its overall score moves ± 6.1 points between runs of the same brief, versus a board median of ± 3.8. A great draft and a mediocre one can come from the identical prompt. The shaded bands on the metric bars below show where that volatility lives.
What it does best is Anti-Slop: it ranks 68th on that dimension at 85.3, above the board average. Its softer spot is Visual Cues (49.4, 118th), which is the thing to watch if that metric matters most for your use.
It was uneven across the 9 scripts. Its best run was the founder announcement / personal origin story (80.3) and its weakest was the personal technical walkthrough / agentic workflow case study (59.5), a spread of about 21 points. Worth knowing if your writing skews toward one of those formats.
On cost, it runs about $0.018/task. DeepSeek V4 Flash (native chat alias) scores higher for less money, at $0.002/task, so on pure value this config is not the frontier. Still, it beats 3 models that cost noticeably more.
Skill profile
The shape is the story: the further a corner reaches, the stronger that dimension. Biggest edge over the board: Anti-Slop (+2.6); furthest behind: Length Discipline (-27.5).
Per-metric scores
Nine writing dimensions, each scored 0–100 and blended by the editorial weight shown. Rank is against all 130 current-ranked models. The thick bar is this model; the thin lines above and below are the current field's best model (darker) and average (lighter) on the same scale. The small ± number is its run-to-run variation.
Per-article scores
The same 9 real scripts every current-ranked model writes, scored individually. Different formats stress different skills.
Consistency: run-to-run variability
Every cell on this page is the mean of 5 independent runs per script; the ± numbers are the run-to-run standard deviation. Overall, Grok 3 varies by ± 6.1 points between runs versus a board median of ± 3.8, so it is swingier than the typical model here, worth knowing if you need repeatable output.
Its most volatile dimension is Visual Cues (± 13.3 vs a board median of ± 4.7): two runs of the same brief can land visibly different visual cues scores.
| Metric | This model | Board median | Verdict |
|---|---|---|---|
| Tone & Voice ? | 73.1 ± 2.9 | ± 2.2 | typical |
| Writing Craft ? | 78.0 ± 2.3 | ± 2.0 | typical |
| Substance & Value ? | 76.9 ± 2.4 | ± 2.6 | typical |
| Flow & Emotion ? | 71.6 ± 2.7 | ± 2.6 | typical |
| YouTube Structure ? | 64.8 ± 5.2 | ± 3.6 | typical |
| Hook ? | 77.6 ± 4.1 | ± 3.0 | typical |
| Length Discipline ? | 39.7 ± 5.5 | ± 8.6 | typical |
| Anti-Slop ? | 85.3 ± 2.6 | ± 2.6 | typical |
| Visual Cues ? | 49.4 ± 13.3 | ± 4.7 | swingier than most |
Measured, not modeled
Numbers we log on every run and rarely talk about. None of these affect the writing scores; cost and latency are informational.
What each judge scored it
The published overall of 70.5 is the consensus of three family-disjoint judges scoring the same 45 stored drafts blind with the identical rubric. The highest and lowest judge differ by 11.1 points on its overall. A typical amount of judge disagreement for this board. How the panel works: methodology.
How we ran Grok 3
- Accessed viaxAI API · model page ↗
- Exact model / route id
grok-3(provider default reasoning) - Generations5 independent runs × 9 scripts = 45 drafts, stored immutably after generation
- Run dates2026-07-29 → 2026-08-11
- Judges3-judge panel: Claude Opus 5, high effort (Anthropic) + DeepSeek V4 Flash (OpenRouter) + GPT-5.6 Sol, medium reasoning (Codex CLI): every draft scored independently by all three with the identical rubric, blind to model identity, evidence quotes required before each metric score; published score = per-metric consensus mean, verified with a Bradley-Terry refit (methodology)
- Samplingprovider-default sampling, no fine-tuning, no system-prompt changes; every current-ranked model gets the identical brief, style contract and research packet
- What stays privatethe exact rubric and any non-public reference material; source-publication timing is disclosed for exceptions (why)
Frequently asked questions
How good is Grok 3 at writing?
On ToneBench it ranks #103 of 130 with a writing Elo of 1016 and an overall score of 70.5/100. That score comes from writing our 9 real YouTube scripts five times each and scoring every draft blind against our own finished versions across nine writing dimensions.
Is Grok 3 the best xAI model for writing?
Not quite. Within xAI it ranks 6th of 7; Grok 4.5 is the family's best writer here.
Is Grok 3 good value for the money?
It costs about $0.018/task. DeepSeek V4 Flash (native chat alias) scores higher for less, so it is not the value pick.
What are Grok 3's strengths and weaknesses?
Its strongest dimension is Anti-Slop (68th on the board, 85.3). Its weakest is Visual Cues (118th on the board, 49.4). The full nine-metric breakdown is on this page.
Does Grok 3 write some formats better than others?
Yes. Its best of our 9 scripts was the founder announcement / personal origin story (80.3) and its weakest was the personal technical walkthrough / agentic workflow case study (59.5).
How consistent is Grok 3 between runs?
We run every script 5 times. Grok 3's overall score varies by about ±6.1 points between runs, versus a board median of ±3.8. That is swingier than typical, so expect more draft-to-draft variation. The full per-metric variability table is on this page.
How was Grok 3 evaluated?
Via xAI API using the exact model/route id grok-3, run on 2026-07-29. 5 runs per script, provider-default sampling, no fine-tuning; every draft scored blind with a fixed rubric by a three-family judge panel. Full details in the 'How we ran it' section and the methodology.
Is Grok 3 open source?
No, it is a proprietary (closed-weights) model. Among proprietary models it ranks 76th of 83 for writing.