Meta's Muse Ember tops small models on Arena's Code Arena WebDev board
Overview
Arena reports that Meta's Muse Ember (Max) scores 1608 points on the Code Arena WebDev leaderboard, the highest of any small model, ahead of Claude Haiku 5.5 (High) at 1595 and GPT-6 Luna (Max) at 1578.
Arena's first report put Ember's median cost per task at $0.06, about half of Luna's and 3.7 times lower than Haiku's. A later Arena post puts the median at $0.03, compared with $0.06 for Luna and $0.11 for Haiku. All three models share the same API price of $0.10 per million input tokens and $0.50 per million output tokens.
In Arena's timed tests, Ember completed 25 complex generations in about 2 minutes per task at a total cost of $1. Meta and Alexandr Wang repeated the ranking and a figure of about $0.04 per website, citing Arena's test of 20+ websites and animations generated in one go.
Written by AI from the articles below · updated Nov 9, 4:26 PM ET
Check the sources:
Developments
2 developments
- Nov 9, 1:56 PM ET · 1 articleCode Arena ranks Meta's Muse Ember highest among small modelsAlexandr Wang: Muse Ember generates websites for about $0.04 each, Meta says
- Nov 9, 1:20 PM ET · 2 articlesArena reports Muse Ember leads small models on Code Arena WebDevArena: Arena ranks Meta's Muse Ember top small model in Code Arena WebDev
Article timeline
The articles in this story. Times are ET.
Alexandr Wang@alexandr_wangXMuse Ember generates websites for about $0.04 each, Meta saysAIMeta says its Muse Ember (Max) model ranks highest among small models in Arena's Code Arena WebDev leaderboard. Arena reports that Ember generated 20+ websites and animations in one go, averaging about 2 minutes and $0.04 per task across timed generations.
Arena.ai@arenaOfficialArena ranks Meta's Muse Ember top small model in Code Arena WebDevAIArena says Meta's Muse Ember (Max) is the highest-performing small model in Code Arena: WebDev, scoring 1608 points. It outperforms Claude Haiku 5.5 (High) at 1595 points and GPT-6 Luna (Max) at 1578 points. Arena puts Ember's median cost at $0.03 per task, compared with $0.06 for Luna and $0.11 for Haiku, with all three sharing $0.10/$0.50 per million input/output tokens.

Arena.ai@arenaOfficialMeta's Muse Ember tops small models on Code Arena WebDev at 1608 pointsAIArena reports that Meta's Muse Ember (Max) scores 1608 points on Code Arena WebDev, the highest of any small model and ahead of Claude Haiku 5.5 (High) at 1595 and GPT-6 Luna (Max) at 1578. Arena says Ember's median cost per task is $0.06, about half Luna's cost and 3.7 times lower than Haiku's, despite identical API pricing of $0.10/$0.50 per million input/output tokens. In Arena's own testing, Ember completed 25 complex generations in about 2 minutes per task for a total of $1.

Heat Trend
- Current comparable heat
- 17
- Peak comparable heat
- 18Nov 9, 5:00 PM ET
- Change in past 24 hours
- –
The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.