Meta's Muse Ember scores 36 on AA Intelligence Index, matching Claude Haiku 5.5 at 1/3 the cost per task
ArtificialAnlys · x · 2026-11-10
Artificial Analysis benchmarked Meta's new efficient-tier model Muse Ember: it scores 36 on the Intelligence Index, matching Claude Haiku 5.5 (max) at roughly a third of the Cost per Task ($0.14 vs $0.41), and about 4x cheaper than DeepSeek V4.1 Flash.
Key results
- Agentic automation: 68.1% on AutomationBench-AA, level with Muse Wonder and GPT-6 Astra; beats Claude Haiku 5.5 on GDP.pdf, trails on AA-Briefcase.
- Agentic coding is mixed: wins Terminal-Bench 4.0 (40.9% vs 32.8%) but loses FrontierCode 2.0 (27.2% vs 34.9%).
- Strong in its own agent: 46 on the Coding Agent Index inside Muse Code, ahead of Claude Haiku 5.5 in Claude Code (37) and GPT-6 Luna in Codex (31).
- Weak knowledge, low hallucination: AA-Omniscience accuracy of 21.1% is the lowest among comparable models, mostly because it declines 64% of questions — keeping hallucination rate at 17.4%, half of Haiku 5.5's.
Pricing: $0.10 input / $0.004 cached input / $0.50 output per 1M tokens; cached input is 60% cheaper and it uses 1/3 the output tokens of Haiku 5.5. Available via Meta's first-party API and Muse Code.
More from Models
- Grok nears 100 million monthly active users — NicoVerderosa · 2026-11-10
- Upgrading a free account still nets you Fable resets, user reports — Sauers_ · 2026-11-10
- Upgrading a free account keeps your Fable weekly resets, user discovers — Sauers_ · 2026-11-10
- Matt Shumer burned 400B+ tokens testing Fable 5.2 — here's what he built — mattshumer_ · 2026-11-10
- Head-to-head: Claude Fable 5.2 beats Opus 5.5 in 4 of 5 real-world tests — petergyang · 2026-11-10
- Ocarina of Time 'gameplay' tweet turns out to be 3D generated by Claude Fable 5.2 — flavioAd · 2026-11-10