Muse Ember
- Runs here
- 6of 6
- List price of its runs
- $0.065
- Total time
- 3 min 48 s
- Price
- $0.50/M out
- Hacker News
- 346pts · 4 stories
Attempts
Pelican riding a bicycle (SVG)
37 s$0.00917k tokensAPI
Wristwatch that shows the current time
40 s$0.01123k tokensAPI
Landing page for a cargo bike workshop
41 s$0.01326k tokensAPI
61 s$0.02039k tokensAPI 
45 s$0.01223k tokensAPI Patchbay is a three-person studio in Rotterdam repairing vintage synthesizers. We work carefully, one instrument at a time, because old machines deserve patience. Our bench is crowded with test leads, spare knobs, tired capacitors, and service manuals held together by tape. Every repair begins with listening. We power up the synth, play a few notes, and trace the fault by ear before opening the case. Perhaps the filter drifts as it warms. Maybe a voice cuts out, or the envelope behaves strangely. We clean contacts, replace components only when necessary, and keep original parts wherever possible. We believe repair is a form of preservation. These instruments were built to be played for decades, not replaced when something small fails. Rotterdam has a long history of making and fixing things, and we are proud to continue that tradition. Bring us your beloved keyboard, and we will see what can be saved.
4 s<$0.0011.9k tokensAPI
About
Meta says Muse Ember is its smallest and most efficient model, built for the work agents repeat all day: tool calls, coding subtasks, routing, extraction and long document parsing. Meta also says it matches Claude Haiku 5.5's best score at about a third of the cost. It lists at $0.10 per million input tokens and $0.50 per million output tokens, with a context of 1,048,576 tokens. The weights are closed.
On HN the top story got 321 points. simonw wrote "There's usually a reasonably strong correlation between thinking level and visual quality of the pelican." Retro_Dev wanted open weights: "I can't find any weights, how sad :(". On X, @rynorhn wrote "meta is so back." (61 likes), and the launch post from @finkd got 9,681 likes.
All six prompts worked on the first try, one shot each, in 3 min 48 s for $0.065. The pelican took 37 s against a median of 75 s over 50 models, but used 17k output tokens against a median of 10k. Breakout took 61 s and 39k output tokens. The pelican is a white bird with an orange pouch on a teal bicycle, with a basket of fish.
Sources: OpenRouter · Launch post · HN thread · @rynorhn on X. Written by Claude Sonnet 5.5 on 10 Nov 2026 from these sources and our runs, then checked by Claude Opus 5.5 and by a script that matches every number and quote to a source.
Specs
- Price in
- $0.10 / M
- Price out
- $0.50 / M
- Context
- 1,048,576 tokens
- Max output
- 943,718 tokens
- Input
- text, image, video, file, audio
- Weights
- closed
- Made in
- US
- Served in
- not listed by its 2 providers
- OpenRouter
- meta/muse-ember
On Hacker News
346 HN pts · 4 stories
Top story: Muse Ember: Our smallest, most efficient model (330 points)
Compare
Also side by side with Nex-N2.6, ricursiveAI/Ripple-65B-Preview, GPT-6.1 Terra, Kimi K3.1 Pro, MiniMax M3.1 Flash, Nemotron 3.5 Super VL, Gemini 4 Argon, LFM2.5-24B-A2B, Muse Wonder, Claude Haiku 5.5, Mistral Large 4, GPT-6.1 Sol, Claude Opus 5.5, GPT-6 Luna, DeepSeek V4.1 Flash, Gemini 3.8 Flash, Claude Fable 5.1, GLM 5.3 Flash, GLM 5.3 and Kimi K3.