ds DeepSeek V4 Flash (native chat alias)
DeepSeek · open weights · writing benchmark
The short version
DeepSeek V4 Flash (native chat alias) sits at #69 of 130 on ToneBench, in the middle of the pack, with a writing Elo of 1490 and an overall score of 77.8 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 DeepSeek it ranks 5th of 8. DeepSeek V4 Flash 0731 is the family's top writer here, about 465 Elo ahead, so this config trades some quality for whatever it saves you in cost or speed.
Among open-weights models, it comes in 14th of 47. Open weights still trail the closed frontier on pure voice fidelity, and you can see that in the gap at the top.
Worth knowing before you rely on it: this model is swingy. Its overall score moves ± 7.8 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 Length Discipline: it ranks 64th on that dimension at 72.9, above the board average. Its softer spot is YouTube Structure (70.8, 89th), 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 product release announcement / personal observation (84.5) and its weakest was the personal technical walkthrough / agentic workflow case study (72.2), a spread of about 12 points. Worth knowing if your writing skews toward one of those formats.
On cost, it runs about $0.002/task. For that money it beats 42 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: Visual Cues (+6.1); furthest behind: Anti-Slop (-3.0).
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, DeepSeek V4 Flash (native chat alias) varies by ± 7.8 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 Tone & Voice (± 4.8 vs a board median of ± 2.2): two runs of the same brief can land visibly different tone & voice scores.
| Metric | This model | Board median | Verdict |
|---|---|---|---|
| Tone & Voice ? | 79.3 ± 4.8 | ± 2.2 | swingier than most |
| Writing Craft ? | 80.0 ± 3.5 | ± 2.0 | swingier than most |
| Substance & Value ? | 79.7 ± 5.3 | ± 2.6 | swingier than most |
| Flow & Emotion ? | 75.8 ± 4.5 | ± 2.6 | swingier than most |
| YouTube Structure ? | 70.8 ± 6.0 | ± 3.6 | swingier than most |
| Hook ? | 81.8 ± 6.5 | ± 3.0 | swingier than most |
| Length Discipline ? | 72.9 ± 9.5 | ± 8.6 | typical |
| Anti-Slop ? | 79.8 ± 4.0 | ± 2.6 | swingier than most |
| Visual Cues ? | 80.6 ± 5.1 | ± 4.7 | typical |
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 77.8 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 8.8 points on its overall. A typical amount of judge disagreement for this board. How the panel works: methodology.
How we ran DeepSeek V4 Flash (native chat alias)
- Accessed viaDeepSeek API · model page ↗
- Exact model / route id
deepseek-chat(provider default reasoning) - Audited route overrideOpenRouter ·
deepseek/deepseek-v4-flashforautomated_video_workflow(Task-7 exact V4 Flash non-thinking route; historical native deepseek-chat alias was formerly mislabeled V3) - Corrected model identityDeepSeek V4 Flash — Formerly mislabeled DeepSeek V3; native alias served V4 Flash for every cached benchmark run
- 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 DeepSeek V4 Flash (native chat alias) at writing?
On ToneBench it ranks #69 of 130 with a writing Elo of 1490 and an overall score of 77.8/100. 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 DeepSeek V4 Flash (native chat alias) the best DeepSeek model for writing?
Not quite. Within DeepSeek it ranks 5th of 8; DeepSeek V4 Flash 0731 is the family's best writer here.
Is DeepSeek V4 Flash (native chat alias) good value for the money?
It costs about $0.002/task. Nothing meaningfully cheaper outscores it, which puts it on the value side of the board.
What are DeepSeek V4 Flash (native chat alias)'s strengths and weaknesses?
Its strongest dimension is Length Discipline (64th on the board, 72.9). Its weakest is YouTube Structure (89th on the board, 70.8). The full nine-metric breakdown is on this page.
Does DeepSeek V4 Flash (native chat alias) write some formats better than others?
Yes. Its best of our 9 scripts was the product release announcement / personal observation (84.5) and its weakest was the personal technical walkthrough / agentic workflow case study (72.2).
How consistent is DeepSeek V4 Flash (native chat alias) between runs?
We run every script 5 times. DeepSeek V4 Flash (native chat alias)'s overall score varies by about ±7.8 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 DeepSeek V4 Flash (native chat alias) evaluated?
Via DeepSeek API using the exact model/route id deepseek-chat, run on 2026-07-29. Effective identity: DeepSeek V4 Flash. Formerly mislabeled DeepSeek V3; native alias served V4 Flash for every cached benchmark run. 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 DeepSeek V4 Flash (native chat alias) open source?
Yes, it is an open-weights model. Among open-weights models it ranks 14th of 47 for writing.