Muse Spark
Meta's large language model, usable via API and coding tools such as OpenCode.
Liked as a cheap, fast model for coding agents123 that some commenters rate near Opus or Sonnet tier and use for checking code against specs.456 The most common caveat is that the cheap contributor tier shares your data with Meta for training,78910 and one user found its generated code and reviews poor, with mostly hallucinated issues.11
Why people recommend it
Coding ability · 3
Commenters say it works well for coding agents, with some calling it close to Opus tier.
“First time I've been impressed by meta's AI products. It feels not far below opus 4.8ish”
“the thing flies, it's a pleasure to use”
Show 1 more
“perfectly capable of following my spec and then poking holes in the implementation”
Price · 3
Commenters cite a very low per-token price and say it is worth it for the cost.
“unbeatable price/intel ratio per M tokens:”
“I used spark 1.3 extensively in prime-agent (a pi derivative) and I thought it was great.”
Show 1 more
“the intelligence gap between DS 4.1 Flash and Muse Spark 1.3 makes Muse more worth it for me”
Free access · 3
Users report free or generous access through OpenCode Zen and OpenCode Go.
“First time I've been impressed by meta's AI products. It feels not far below opus 4.8ish”
Show 1 more
“its free on opencode and i use it for personal projects”
Reasoning and vision · 3
Testers say it solved a floor-plan PDF task and common gotcha prompts that other models missed.
“it knocked it out of the park, extracted and displayed the pertinent pages from a multi-page PDF inline in the chat and rendered a correct answer.”
“tried it (with instant mode), it's an impressive model.”
Show 1 more
“This makes it a very interesting alternative to Deepseek for personal work where I don't care about the training”
Speed · 2
Users describe it as very fast, at over 200 tokens per second.
“the thing flies, it's a pleasure to use”
Writing Latin · 1
One learner says it is among the best models for writing Latin.
“I find it's one of the best at it, alongside Gemini.”
Caveats and dissent
Data training · 3
The cheap contributor tier shares your data with Meta for training.
“the thing flies, it's a pleasure to use”
“the intelligence gap between DS 4.1 Flash and Muse Spark 1.3 makes Muse more worth it for me”
Show 1 more
“This makes it a very interesting alternative to Deepseek for personal work where I don't care about the training”
Code quality · 3
Some users rank it below frontier models or DeepSeek Flash, and one found its reviews were mostly hallucinated.
“the code it generated, while technically working, was entirely garbage”
“I used spark 1.3 extensively in prime-agent (a pi derivative) and I thought it was great.”
Show 1 more
Debugging loops · 1
Version 1.2 could get stuck repeating thought summaries without progress on bugs.
Integration quirks · 1
Its server-side tool calling caused a bug when used with Codex.
“also a very cool model that I'm enjoying using so far!”
Meta account · 1
One commenter disliked needing a Facebook account for Meta's coding products.
“tried it (with instant mode), it's an impressive model.”
Often compared with
1 moved from Muse Spark to GLM Flash
1 moved from Claude to Muse Spark
Who recommends it in their own words
Threads where it came up
- Muse Spark 1.3 6 mentions
- Muse Spark Open 2 mentions
- GPT-6 Sol and Luna 2 mentions
- Ask HN: What default model do you use and why? 2 mentions
- Muse – Meta’s personal AI agent 2 mentions
- Muse Code and Muse Spark 1.2 2 mentions
Also written as Muse Spark 1.3, Muse Spark 1.2. Data: muse-spark.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 23
“I think Muse Spark 1.3 is better but GLM 5.3 Flash is an excellent daily driver.”
“I used spark 1.3 extensively in prime-agent (a pi derivative) and I thought it was great.”
“I'm not sure I'll be compelled enough to switch from Muse Spark 1.3”
“the intelligence gap between DS 4.1 Flash and Muse Spark 1.3 makes Muse more worth it for me”
“the code it generated, while technically working, was entirely garbage”
“perfectly capable of following my spec and then poking holes in the implementation”
“Adding Muse Spark 1.3 to that mix but not convinced yet.”
“unbeatable price/intel ratio per M tokens:”
“First time I've been impressed by meta's AI products. It feels not far below opus 4.8ish”
“the thing flies, it's a pleasure to use”
“I used 1.2 for free for a while, and it was a pretty good experience.”
“You can check out Muse Spark 1.3 by using OpenCode”
“I find it's one of the best at it, alongside Gemini.”
“I use it through Opencode Go”
“its free on opencode and i use it for personal projects”
“it has been quite good and fast”
“I would rank MS 1.2 below it.”
“Meta is rocking AI. As of last week I have been using their excellent muse coding harness with their model Muse Spark 1.2.”
“it would be nice to know how it compares on a cost/performance basis against models in its class.”
“This makes it a very interesting alternative to Deepseek for personal work where I don't care about the training”
“also a very cool model that I'm enjoying using so far!”
“it knocked it out of the park, extracted and displayed the pertinent pages from a multi-page PDF inline in the chat and rendered a correct answer.”
“tried it (with instant mode), it's an impressive model.”