Breaking

OpenAI GPT-6 Astra Max takes first on Code Arena WebDev at 1,797

September 5, 2026 at 13:57 EDT

Software Dev & Coding
Foundation Models
AI Agents

OpenAI’s GPT-6 Astra, scored at Max reasoning as gpt-6-astra-max, is first on Code Arena’s WebDev leaderboard at 1,797 points, 35 ahead of Anthropic’s Claude Fable 5.1 Max, two days after the model began rolling out.

Sep 5, 2026 · Code Arena

Astra Max takes WebDev

OpenAI’s gpt-6-astra-max leads Code Arena WebDev at 1,797 points — 35 ahead of Claude Fable 5.1 Max, at the same $40/M blended price.

WebDev overall (spread 1–2)

1797

Arena score (±24)

+180

pts vs Sol xHigh (#13)

$40/M

blended; list $10 / $50

WebDev Arena score (same scale)

Column height is proportional to the published score. Close heights mean a tight race.

1797

Astra Max
OpenAI

1762

Fable 5.1 Max
Anthropic

1688

Opus 5 Max
Anthropic

1617

Sol xHigh
OpenAI, #13

List price vs GPT-5.6 Sol (input, per 1M tokens)

Astra is 2.5× Sol on the rate card ($10 vs $4). Token efficiency can narrow cost per finished coding task.

Sol

$10

Astra

How Code Arena scores a model

1 Prompt → 2 Tool-using agent → 3 Live web app → 4 Human vote

What improved

First on WebDev overall; +180 vs Sol xHigh; token-efficient on independent coding-agent tests; early #1 in Data & Analytics, Consumer Product, and Content Creation Tools.

What still constrains it

Rank spread 1–2 on fewer votes; Fable 5.1 leads some coding indexes and Simulations (1,873 vs 1,821); 2.5× Sol list price; reasoning cannot be turned off; UI taste still debated.

Continue reading

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