Muse Spark
Muse SparkAlternatives?
CheaperGemini 3.8 Flashunstated+3%-25%*$0.75 / $3.75Sampling parameters are ignored+1
BetterClaude Fable 5unstated+5%+900%$10 / $50
OpenKimi K3unstated+2%+200%$3 / $15Sampling parameters are fixed+3CheaperGemini 3 Flash Previewunstated+3%-44%$0.50 / $3CheaperGemini 3.6 Flashunstated+3%-25%*$0.75 / $3.75Gemini 3.8 Flashunstated+3%-25%*$0.75 / $3.75Sampling parameters are ignored+1Gemini 3 Flash Previewunstated+3%-44%$0.50 / $3Gemini 3.6 Flashunstated+3%-25%*$0.75 / $3.75
MiniMax M3unstatedOpen+1%-74%$0.30 / $1.20Context: 1M → 524K+2
Qwen 3.5 Plusunstatedsame-87%$0.12 / $0.69Claude Fable 5unstated+5%+900%$10 / $50Claude Fable 5.1unstated+5%+900%$10 / $50Replace forced tool_choice+5Gemini 4 Argonunstated+5%+100%*$2 / $10Synchronous output stops at 65,536+2Gemini 3.1 Pro Previewunstated+4%+125%$2 / $12Cannot use both thinking controlsGemini 3.8 Flashunstated+3%-25%*$0.75 / $3.75Sampling parameters are ignored+1Kimi K3unstatedOpen+2%+200%$3 / $15Sampling parameters are fixed+3MiniMax M3unstatedOpen+1%-74%$0.30 / $1.20Context: 1M → 524K+2
GLM 5.3unstatedOpensame+7%$1.40 / $4.40No image input+4Kimi K2.6unstatedOpensame-14%$0.95 / $4Context: 1M → 256KGLM 5.1unstatedOpensame+7%$1.40 / $4.40Context: 1M → 203K+1Muse Spark 1.2unstated+1%—$1.25 / $4.25Only auto tool_choice is supported+1Muse Spark 1.1unstatedsame—$1.25 / $4.25Only n=1 · Meta / Chat Completions+1Llama 4 MaverickunstatedOpen-8%—$0.19 / $0.65Llama 4 ScoutunstatedOpen-14%—$0.10 / $0.30Muse Spark OpenxhighOpenn/a—$0.84Reasoning off leaves output unchanged · Self-hosted / llama.cpp+1Muse Wondermaxn/a—$4.52Only auto tool_choice is supported+5Muse Wonder Fulln/a——Only in Muse CodeMuse Spark 1.3xhighn/a—$1.37Only auto tool_choice is supported+5All Meta models →History
4 versionsBenchmarks
in Chat & KnowledgeWatch out 2
Only n=1 · Meta / Chat Completions›
Requests with n greater than 1 return HTTP 400; request one completion at a time. Source ↗
Chat Completions cannot carry reasoning · Meta / Chat Completions›
For external keys, Chat Completions redacts reasoning_content to empty and cannot carry reasoning across turns; use Responses when continuity matters. Source ↗
FAQ
How much does Muse Spark cost?›
$1.25 per 1M input tokens and $4.25 per 1M output tokens; cached input $0.15 per 1M.
Is Muse Spark free to use?›
There is no free API; the cheapest paid price is $1.25 per 1M input tokens.
How good is Muse Spark, and what is it best at?›
#42 of 334 in Chat & Knowledge, #3 of 19 in Coding, #79 of 175 in Agents & Tool Use, #15 of 195 in Writing & Language, #16 of 124 in Vision & Documents. Strongest at: Tool calling, Reasoning, Science, Long context, Knowledge, Instructions, Vision.
Is Muse Spark better than Qwen3.6 Plus?›
On Vals LegalBench, Muse Spark scores 84.2 vs 84.2 — about the same.
How can I use Muse Spark?›
Through the API of Meta.
What is the context window of Muse Spark?›
1M tokens.
When was Muse Spark released? Is there a newer version?›
Released 2026-04-08. It has been superseded by Muse Spark 1.1.
Is Muse Spark open source?›
No, Muse Spark is a closed model available through an API. The closest open-weights option here is Kimi K3.
Can Muse Spark read images?›
It accepts text and images, audio as input.
What is Muse Spark?›
Muse Spark is a closed language model from Meta, released Apr 2026.
What are the best alternatives to Muse Spark?›
Gemini 3.8 Flash (cheaper), Claude Fable 5 (better), Kimi K3 (open weights).