Muse Spark 1.3
Rank of 27063from #3
Known for
Long-horizon agentic coding on Muse Code or the Meta Model API, at a workhorse price.
Meta's newest Spark, out on 2 September 2026 and level with GPT-5.6 Sol on Terminal-Bench 2.1 at a third of the price. Trained for agentic workflows, and the first Spark to lead a coding table outright on DeepSWE.
What it does8
- 1M-token context
- Text, image, video and document in
- Fewer tool calls and fewer tokens on the same coding tasks than 1.2
- Keeps long-form constraints across multi-step work
- Corrects gaps in its own plan mid-task
- Asks before an irreversible action
- Stronger resistance to prompt injection
- Same $1.25 / $4.25 standard rate as 1.2, with a contributor tier at $0.10 / $0.20
From Muse Spark 1.3’s own site. Not used for ranking.
68index score · LLMs
How well does it score?
Compared with everyone else on this board.
—votes · crowd
Reading votes…
Votes don’t change the score.