Breaking

Meta's Muse Ember (Max) Leads Small Models on Code Arena WebDev Leaderboard

November 9, 2026 at 13:27 EST

Foundation Models
Software Dev & Coding

Meta's Muse Ember (Max) has been named the highest-performing small model on the Code Arena WebDev leaderboard, scoring 1608 points and ahead of two leading small models from Anthropic and OpenAI.

Code Arena · WebDev leaderboard · Small models

Meta's Muse Ember (Max) takes the top small-model spot on WebDev

Muse Ember (Max) scored 1608 points, edging out Claude Haiku 5.5 (High) and GPT-6 Luna (Max) in web development tasks. The lead is narrow, but it places Meta's model first among small models on the board.

1608

Muse Ember (Max) points

1595

Claude Haiku 5.5 (High) points, ranked #35

1578

GPT-6 Luna (Max) points, ranked #42

Points on the WebDev leaderboard

Each column's height shows points above a 1500 baseline (2 px per point). The axis is truncated, so the gaps look larger than the raw scores.

1608

Muse Ember (Max)

1595

Claude Haiku 5.5 (High)

1578

GPT-6 Luna (Max)

Size of the margin, in points

One block = one point. Muse Ember's lead over each rival, drawn to scale.

13 pts

vs Claude Haiku 5.5 (High)

Continue reading

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