Claude Opus 4.8

Anthropic, closed weights, released May 2026, run at provider defaults through the Anthropic API. 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

52.9

95% interval 48.0–60.5

Rank: 14of 14 models by Index; the paired intervals allow 12 to 14
Cost per page: $0.15Generation only, list price on run day
Second judge: 69.5GPT-5.6 Terra places it 14th of 14. The Claude judge: 14th.
Grade sum: 21.80Seven 1–5 grades, out of 35
Pages judged: 1520 judge passes, 3 builds per brief
Defects per page: 2.5Content wider than the phone (9 of 15)

Rank 14 of 14; the data allows anywhere from 12 to 14. Within noise of Gemini 3.8 Flash and DeepSeek V4 Pro: 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, Muse Wonder, Claude Opus 5.5, GPT-5.6 Sol, Claude Fable 5.1 and Claude Fable 5. Not clearly ahead of any model.

Designer check pending · read Nov 6Season 0.5Provisional: 1.3 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
Color3.0514 of 14
Components3.0214 of 14
Composure3.1712 of 14
Craft3.1714 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.

Landing51.514 of 14
Dashboard48.814 of 14
Settings47.313 of 14
Pricing49.113 of 14
Article67.612 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: 61sMedian over every build, broken ones included
Output tokens: 6.2kMedian per page, reasoning included
Reasoning tokens: n/aIts API doesn’t report them separately
Broken pages: 0 of 15Every build came back as a whole page
Criticals: 33%5 of 15 scored pages had a critical issue in their first Rams code review
Built through: Anthropic APIServed by Anthropic

Best and worst page

Of its 15 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: Article73.2
Index on this page, one judge pass. Brief mean 67.6.
- Drop cap 'E' visually heavy but unrefinedModerate
- Blockquote left border creates slightly tight internal paddingModerate
- Cite attribution in pullquote competes with quote weightModerate
Worst: Settings45.5
Index on this page, mean of 2 judge passes. Brief mean 47.3.
- Inconsistent button styles across sectionsSerious
- Inconsistent button treatment across same-level actionsSerious
- Avatar block and form fields misaligned vertically in profile rowSerious

All 15 of its pages sit beside every other model's in the gallery.

Head to head

Claude Opus 4.8 against every other model: the paired difference in Index points, and whether the 95% interval on it clears zero.

vs GPT-6 Astra-13.6Clearly behind
vs Grok 4.7-13.2Clearly behind
vs GPT-6.1 Sol-12.5Clearly behind
vs Claude Sonnet 5.5-8.9Clearly behind
vs GPT-6 Luna-8.9Clearly behind
vs Gemini 4 Argon-8.8Clearly behind
vs Muse Wonder-7.7Clearly behind
vs Claude Opus 5.5-7.5Clearly behind
vs GPT-5.6 Sol-6.1Clearly behind
vs Claude Fable 5.1-6.1Clearly behind
vs Claude Fable 5-4.8Clearly behind
vs Gemini 3.8 Flash-2.7Inside the noise
vs DeepSeek V4 Pro-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