Muse Ember

Scores

Dimension	Score	Rank	Benchmarks
Overall: Balanced	61.8	17	10
Overall: Coding	61.2	20	9
Overall: Reasoning	62.0	21	2
Overall: Agentic work	61.9	19	7
Overall: Frontend	60.0	20	2
Agents	62.0	16	6
Reasoning	61.6	31	1
Knowledge	62.7	13	1
Frontend	60.0	21	2
Value vs price	+18.9	1

🚀 What a week for AI! Qwen4-Max enters preview, while Muse Ember, Fable 5.2 and Gemini 4 Carbon have all landed. @elonmusk says Grok 4.8 gets a soft rollout this week after xAI rebuilt its RL/inference stack in C/C++. OpenAI, GPT-6.2 Sol and Astra better be worth the wait! …@RealCamV · 7 likes
Gamefeel RE subsection: Muse Ember and Muse wonder max added they perform so poorly and it was so expensive to run that I dont feel like there is any point running all the other subsections https://t.co/A4kmlJohtV@follistatindev · 3 likes · negative
Nemo Nautilus is the LLM Infrastructure Engineer and testing lead at SMF Works. He gave two AI models the same frozen prompt to build Portofino in Blender, then rendered exactly what they wrote (Blender 4.5.3, Cycles CPU, 64 samples, 1280x720, the first 12 seconds of each camera…@MichaelGannotti · 3 likes

Configuration	Balanced	Benchmarks
Unspecified setting	61.9 ±5.0	2
Max	61.9 ±4.6	8

Benchmark	Result	Source	Date
Terminal-Bench 4.0 (Vals run)	39.4%	Vals AI	2026-11-09
APEX-Agents	64.3%	Epoch AI	2026-11-11
EMB v1.1 (Excel financial modeling)	72.2%	Vals AI	2026-11-09
Tax Agent Bench	38.2%	Vals AI	2026-11-09
Finance Agent v2.1	63.8%	Vals AI	2026-11-09
HLAB-Vals 1.1 (legal agent)	25.0%	Vals AI	2026-11-09

Benchmark	Result	Source	Date
SimpleBench	72.7%	SimpleBench	2026-11-10

Benchmark	Result	Source	Date
Legal Research Bench	49.0%	Vals AI	2026-11-09

Benchmark	Result	Source	Date
Vibe Code Bench v1.1	83.7%	Vals AI	2026-11-09
LMArena WebDev	1,608	LMArena	2026-11-09

muse-ember, muse-ember-1-0