DeepSeek V4 Pro

DeepSeek, open weights, released Aug 2026, run at provider defaults through OpenRouter, served by Alibaba. It built the same five pages as every model on season 0.5 of the Rams Index, and Rams graded each one at desktop and phone size.

Rams Index, 0 to 100

54.6

95% interval 48.9–61.4

Rank: 13of 14 models by Index; the paired intervals allow 8 to 14
Cost per page: $0.032Generation only, list price on run day
Second judge: 69.8GPT-5.6 Terra places it 12th of 14. The Claude judge: 13th.
Grade sum: 22.30Seven 1–5 grades, out of 35
Pages judged: 1010 judge passes, 2 builds per brief
Defects per page: 3.2Tap targets under 24px on the phone (7 of 10)

Rank 13 of 14; the data allows anywhere from 8 to 14. Within noise of Muse Wonder, Claude Opus 5.5, Claude Fable 5.1, Claude Fable 5, Gemini 3.8 Flash and Claude Opus 4.8: the gap to each sits inside the noise, so the board can't order them yet. Clearly behind GPT-6 Astra, Grok 4.7, GPT-6.1 Sol, Claude Sonnet 5.5, GPT-6 Luna, Gemini 4 Argon and GPT-5.6 Sol. Not clearly ahead of any model.

Designer check pending · read Nov 6Season 0.5New this seasonProvisional: 1.0 judge passes per page, protocol 2

Grades by dimension

The seven grades behind the Index, 1 to 5, mean of desktop and phone. The tick is the mean across all 14 models.

Hierarchy3.2512 of 14
Typography3.1013 of 14
Spacing3.0513 of 14
Color3.2013 of 14
Components3.3510 of 14
Composure3.1513 of 14
Craft3.2013 of 14

Bars start at 2.8 and end at 4.2, the range the board uses, so small differences show.

Index by brief

The same Index on each of the five briefs, and where the model places on that brief among all 14.

Landing59.812 of 14
Dashboard49.113 of 14
Settings48.212 of 14
Pricing50.98 of 14
Article65.213 of 14

Bars run from 40 to 90.

How it ran

Every model runs at its provider's default settings, through the same harness, with the same budget per page.

Effort: DefaultThe provider’s default. We send no effort or thinking setting
Time per page: 196sMedian over every build, broken ones included
Output tokens: 15.9kMedian per page, reasoning included
Reasoning tokens: 5.8kMedian per page, as the API reports them
Broken pages: 0 of 10Every build came back as a whole page
Criticals: 10%1 of 10 scored pages had a critical issue in their first Rams code review
Built through: OpenRouterServed by Alibaba

Best and worst page

Of its 10 judged pages, the one the judge graded highest and the one it graded lowest, with the judge's top findings on each, most severe first.

Best: Article69.6
Index on this page, one judge pass. Brief mean 65.2.
- Kicker label all-caps tracking competes with byline registerModerate
- Uneven vertical rhythm between article sectionsModerate
- Pull-quote left border accent absent or invisibleModerate
Worst: Pricing46.4
Index on this page, one judge pass. Brief mean 50.9.
- Pricing card CTA vertical misalignmentSerious
- Billing note redundancy creates competing signalsSerious
- Large vertical gap between price and billing detailSerious

All 10 of its pages sit beside every other model's in the gallery.

Head to head

DeepSeek V4 Pro against every other model: the paired difference in Index points, and whether the 95% interval on it clears zero.

vs GPT-6 Astra-11.8Clearly behind
vs Grok 4.7-11.4Clearly behind
vs GPT-6.1 Sol-10.7Clearly behind
vs Claude Sonnet 5.5-7.1Clearly behind
vs GPT-6 Luna-7.1Clearly behind
vs Gemini 4 Argon-7.0Clearly behind
vs Muse Wonder-5.9Inside the noise
vs Claude Opus 5.5-5.7Inside the noise
vs GPT-5.6 Sol-4.3Clearly behind
vs Claude Fable 5.1-4.3Inside the noise
vs Claude Fable 5-3.0Inside the noise
vs Gemini 3.8 Flash-0.9Inside the noise
vs Claude Opus 4.8+1.8Inside the noise

The other models

Season 0.5 of the Rams Index, in board order, then the unranked.

Rams Index v0.5 · updated 2026-10-17Judged by Rams eyes · Claude Sonnet 4.6