Providers/Meta
Meta
Muse Spark — the coding-optimized frontier model from Meta Superintelligence Labs, with a million tokens of context.
Models on Brainbase
About
Meta Superintelligence Labs builds Muse Spark, a coding-optimized model with a one-million-token context window, visual input, and reliable tool calling — built for long agent sessions that keep repository state across turns.
On Brainbase, Muse Spark is platform-billed ($1.25 in · $4.25 out per 1M tokens, $0.15 cached input) and runs inside any open harness — Codex Agent, OpenCode, or Qwen Code — selected per task like any other platform model.
Available on Brainbase:
- Muse Spark: Meta Superintelligence Labs' first model family; the Brainbase id tracks the current 1.2 release, including Muse Spark 1.2
Run these models in Codex, OpenCode, or Qwen Code — selected per task like any other platform model.
Try it on Brainbase
One API call boots a managed sandbox on the opencode harness pointed at meta/muse-spark-1.2. Hosting, scaling, and billing are handled for you.
curl -X POST https://api.brainbaselabs.com/v2/agents \
-H "Authorization: Bearer $BRAINBASE_API_KEY" \
-d '{
"title": "my-agent",
"runtime_kind": "opencode",
"default_model": "meta/muse-spark-1.2"
}'