Breaking

Meta's Muse Spark 1.3 Re-enters Code Arena: WebDev's Top 10 at 1,623 Points

September 3, 2026 at 23:12 EDT

Software Dev & Coding
AI Agents
Foundation Models

Meta's Muse Spark 1.3 has returned to the top 10 of Code Arena: WebDev's early leaderboard, scoring 1,623 AutoEval points at approximately No. 10. In the same snapshot, it sat five points behind Claude Fable 5 and seven points ahead of GPT-5.6 Sol.

September 2, 2026 · Meta

Muse Spark 1.3 returns to WebDev's top 10

An early AutoEval result places Meta's coding model at about No. 10 with 1,623 points, close to Claude Fable 5 and GPT-5.6 Sol.

Fewer moves in coding workflows

Meta's internal comparison reports about 20% fewer tool calls and 25% fewer tokens than Muse Spark 1.2.

Tool calls

100%Muse Spark 1.2

80%Muse Spark 1.3

Token use

100%Muse Spark 1.2

75%Muse Spark 1.3

The signal in four numbers

~#10

Early WebDev rank

1,623

AutoEval points

Token context window

Max preview index

Early feedback splits

What worked

Fast output and clean, functional UI in one hands-on test.

Better SVG detail than Muse Spark 1.2 in another test.

What to watch

Other developers reported web-app mistakes with 1.2 and mixed coding reliability.

Contributor pricing trades cost for permission to use data for training.

Top-10 signal, not final verdict

The early score is promising, but its meaning depends on evaluation method, reasoning availability, and real-world reliability.

EARLY AUTOEVAL · MAX PENDING SAFETY TESTING · LIVE VOTES CAN MOVE THE SCORE

Continue reading

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