Breaking

Claude Sonnet 5.5 xHigh Reaches No. 3 on Code Arena WebDev, Two Points Behind GPT-6 Astra

October 1, 2026 at 12:52 EDT

Software Dev & Coding
Foundation Models
AI Agents

Claude Sonnet 5.5 with xhigh reasoning has reached No. 3 on Arena's Code Arena: WebDev leaderboard, posting 1,786 points and landing only two points behind GPT-6 Astra Max. The placement puts Anthropic's cheaper Sonnet-tier model close to the top of a human-preference web development benchmark that emphasizes multi-step coding workflows and tool use.

Oct. 1, 2026 · Code Arena: WebDev

Claude Sonnet 5.5 closes in on the WebDev leaders

The xHigh run reached No. 3 with 1,786 points, just 2 points behind GPT-6 Astra and 29 points behind Claude Opus 5.5 Max.

Arena rank

No. 3

overall WebDev

Score

1,786

+/- 18 points

Model votes

1,531

for this run

Raw API price

$2 / $10

input / output per 1M

The gap is tiny to No. 2, larger to No. 1

Stacked blocks show point gaps from Claude Sonnet 5.5 xHigh. Each block equals one Code Arena point.

Why builders care

The run puts a Sonnet-priced model near the top of web app generation, with strong signals for design-heavy and multi-step coding prompts.

Why caution remains

The 2-point difference is inside the published uncertainty range, and xHigh can add startup latency and token use.

Effort setting is the decision lever

Default API

high

Apps and Claude Code

medium

xHigh and max

use after evals show a gain

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.