Meta · Model release
Muse Spark 1.3: Meta's self-reported DeepSWE result
In its 2026-09-02 release, Meta reported 75.4% on DeepSWE 1.1 at $1.25/$4.25 per MTok; this is a vendor-reported result, not a universal leaderboard ranking. The same announcement says max reasoning is available on Muse Code and Meta Model API.
What changed
- Meta's chart reports 75.4% DeepSWE1.1 at max versus 55.0% for Spark1.2 at xhigh: a 20.4-point difference across those configurations, not an effort-matched gain. Terminal-Bench2.1 is 88.8%; its methods use native coding harnesses.
- First-party API pricing of $1.25/$4.25 per MTok (cached $0.15) plus a $0.10/$0.20 contributor tier; 1M-token context.
- About 20% fewer tool calls and ~25% fewer tokens than Spark 1.2 in Meta's engineer comparisons.
Best fit
- Long-horizon coding agents
- Frontier quality below frontier pricing
- Meta ecosystem teams
What to be careful about
- Meta's current announcement says max reasoning is available on Muse Code and Meta Model API; confirm the exact mode and eligibility on your account.
- Meta says open weights are on its roadmap but had not released them in the source checked for this guide. Datacurve's live DeepSWE 1.1 snapshot includes a separate Muse Spark 1.3 xHigh row; do not treat it as the same run/configuration as Meta's reported 75.4% result.
Dataset snapshot: Muse Spark 1.3 is also represented in the checked comparison dataset. Prices and benchmark figures carry their own field dates.