Meta Muse Spark 1.3 hits #2 on ProgramBench, full eval data open-sourced
jyangballin · x · 2026-09-30
ProgramBench's author shares eval profiles for Meta Muse Spark 1.3: on the benchmark where a SWE-agent writes whole programs (sqlite, ffmpeg, php) from scratch, 1.3 (max) ranks #2 overall with 2.5% fully resolved and 25.0% almost resolved; 1.3 (xhigh) ranks #3.
Cost: the max tier consumed $1,293 in API spend and 65,037 LLM calls across 200 tasks ($6.5/task). Language breakdown: 58% of C projects were rewritten in Python, while Rust projects stayed in Rust 58% of the time. Trajectories, final codebases, and test pass/fail breakdowns are all open-sourced.
Related event: Meta Muse Spark 1.3 Ranks Second on ProgramBench with Open Evaluation Data(2 posts)→
More from coding & agent
- Generic API-catalog MCP server vs one tool per job: a context-bloat experiment — MandeepKaur9407 · 2026-11-16
- Berkeley paper: LLMs answer with stale values even when the update is in context — ombre-ai · 2026-11-16
- Effective HTML: 4.2k-star agent skills turn wireframes into self-contained interactive prototypes — tom_doerr · 2026-11-16
- skills-gate: one gate skill fronts 148 agent skills at just 389 tokens — Distinct-Pie2389 · 2026-11-16
- Open-weight Karpo-Decision-31B tops Decision Index at 73.9, one forward pass per answer — CodeByPoonam · 2026-11-16
- shadcn boosts @uisdk: one-shot UI rebuild that's instantly E2E-testable by AI — shadcn · 2026-11-16