vibecodingnav
中/EN
Scenarios/web-frontend

Build a Web frontend

Pick the best frontend-writing model from human blind votes, design restore, framework tasks, engineering judgment, long-horizon engineering, and mergeability.

Right now Claude Opus 5.5 is first on 4 of 6 boards it enters (1 tied).

Sorted by compositeSynced 2026-10-10
Claude Opus 5.5976/7112—211
Claude Sonnet 5.5914/7321—4——
GPT-6 Astra876/7235138—
Claude Fable 5.1816/7644—743
GPT-6.1 Sol814/7553—5——
GPT 6 Sol714/7127——62—
Claude Opus 5715/7106—311—11
Gemini 4 Carbon701/74——————
GPT-6.1 Terra672/714————5—
GPT 5.6 Sol665/73112—410—4
Claude Fable 5664/72610—5——7
Gemini 3.7 Flash642/735——12———
Grok 4.6632/730——9———
Qwen3.8 Max634/7139616———
Gemini 4 Argon634/711—8—123—
Muse Wonder633/77—8—13——
Gemini 3.8 Flash625/73719152—7—
Qwen3.8 Flash Next611/724——————
Kimi K3.1 Pro612/79—6————
MiMo V2.6 Flash602/72218—————

Suggested build

Why it matters

Closest to “does it look finished”

Frontend quality is hard to catch with tests. A blind vote is the person's first look: layout, detail, and whether it runs.

New, fast, and large

952,039 votes and 153 models. New models usually show up within days, so this is the freshest frontend signal.

Read the price with the score

Third place, Claude Sonnet 5.5, is 40 points behind first at half the price. Open-weight DeepSeek V4.1 Flash (20261017) is 16th at $0.30 / $1.20.

Other scenarios