Skip to content
OSMEXO
Sign inGet an API key
☰
Sign inGet an API key

Muse Spark Open

meta/muse-spark-open
  • Meta
  • Family · Muse Spark Open
  • Released 15 Oct 2026
  • Open weights
  • Text · Image · Video · File · Audio → Text
  • Reasoning · adjustable
  • Tools · JSON · Vision
Not on OSMEXO yet
Request this model
Spec & list price
Context1M
Max output944K
Input / 1M$1.25
Output / 1M$4.253.4× input
Scores · standing
Intelligence38.9#40/490
LiveBench51.6#26/32
Usage
Tokens / day3.1B▼ 63% 7d
Usage rank#140
Requests / day58.6K
Tool calls / req0.57
Per request · 30 days
Avg input45.5K3.1× median
Avg output1.2K1.8× median
Cache hit66%about median
Reasoning share71%1.3× median
Tool-call failures1%0.5× median

Consumption

2.7B5.3B8.0B26 Oct29 Oct1 Nov
Peak 40.2B 17 OctLow 1.3B 1 NovRank today #140

News & trends

16 mentions this week ▲ 220%this model · videos, posts, articles · 30 days
Trending · 14 days
News

Benchmarks

bar = share of the top score · Benchmark data: Artificial Analysis
GPQA Diamond#56 of 445
89.0%
Long-context reasoning#70 of 393
78.3%
SciCode#18 of 160
58.4%
Terminal-Bench 2.1#32 of 182
76.4%
Humanity's Last Exam#30 of 465
43.9%
Intelligence index#40 of 490
38.9
Terminal-Bench 4.0#50 of 153
8.1%
Speed · median across 6 providers, latest day
Output speed#40/334132 tok/s2.0× faster than median
Fastest provider182 tok/sbest of 6 providers
First token#760.56 s1.8× quicker than median
First answer token#4511 sthinking included

#9 cheapest of 18 with a similar score

Value
Intelligence Index
3638404244$1$2$5Muse Spark Open
Blended price per 1M tokens (log)
LiveBench suites
Browser26.2
Coding73.6
Data Analysis60.2
Design60.3
Long Agentic11.7
Reasoning73.7
Research80.3

▲ Strongest: Research, Reasoning. ▼ Weakest: Long Agentic (12).

Technical details

Reasoning
on · medium
Tokenizer
Other
License
other
Parameters
822B
Moderation
none
Downloads · likes
9.9K · 737
Reasoning effort
  • xhigh
  • high
  • medium · default
  • low
Request parameters 20
  • tools
  • tool_choice
  • response_format
  • structured_outputs
  • reasoning
  • reasoning_effort
  • include_reasoning
  • temperature
  • top_p
  • top_k
+10 more
  • min_p
  • max_tokens
  • stop
  • seed
  • frequency_penalty
  • logit_bias
  • logprobs
  • presence_penalty
  • repetition_penalty
  • top_logprobs

Interesting facts

  1. 9.8× the output room of similar models (943.7K against a 96.8K median).
  2. Released 18 days ago.

When to use

Best for

  • Self-hosted multimodal assistants (open weights)
  • Research and analysis over very long documents
  • Science question answering
  • Audio, video and image understanding

Think twice for

  • Long autonomous agent runs (weakest suite)
  • Browser-driving agents
  • Hardest coding agent work

Alternatives

other vendors · similar capability

Want Muse Spark Open on OSMEXO?

Sign in to request it. Requests are reviewed; a request does not guarantee availability.

Request this model on OSMEXO