SYLIN

Muse Ember

by Meta · Model · Language & code models

promising

Cheap, fast API model that holds up well in cost-versus-quality tests.

Lightweight multimodal Meta model for speed and cost; API and Muse Code harness only, no open weights expected

AccessAPIPrice$0.1 in · $0.5 out per M tokensContext1,049K tokens

Why we say so

Arena measured it about 4x cheaper per task than GPT-6 Luna and Haiku 5.5, and 12x faster than Luna. Vals found it reaches 85% of Claude Fable 5.2's performance at under 1% of the cost. Rowan Cheung reports it ties Haiku 5.5 on the Intelligence Index. The per-token price matches Luna and Haiku, so the saving comes from using fewer tokens. No one has reported hands-on use yet.

Evidence

  1. Nov 15
    +1.7% on Agent Arena at about $0.02 per task: ~4x cheaper than GPT-6 Luna and Haiku 5.5, ~12x faster than Luna
  2. Nov 12
    Reaches 85% of Fable 5.2's performance at 0.84% of the cost
  3. Nov 11
    Reportedly ties Haiku 5.5 on the Intelligence Index at about one-third the cost; one user sees about 400 tok/s

Cost to run

Alternatives

Gemini HydrogenGoogle pitches it against Meta's Muse Ember and likely Claude Haiku 5.5 at its price tier