Praxist Agent Achieves High Cost-Efficiency on MLE-bench
iamfakhrealam · x · 2026-08-28
Praxist is an agent system that keeps all attempts instead of just the winner. It runs multiple research directions in parallel, stores findings in shared memory, and uses a PI panel to select next steps. On all 75 MLE-bench tasks, it achieved a 65.3% gold rate for $3,054 using an open-source model, while the Claude Code + Opus 4.8 baseline got 44% fewer golds for $38,370.
More from coding & agent
- Vercel Goes Full Backend: Python GA, Containers, and Daemons — cramforce · 2026-10-16
- WorldMemArena Accepted at NeurIPS: Memory Storage ≠ Memory Use — xwang_lk · 2026-10-16
- Vercel Ship Demos Container Deploys for Go, Ruby, Java and More — edwin · 2026-10-16
- TRL adds day-0 support for fine-tuning Muse Spark Open's 816B MoE with LoRA on 64 H100s — QGallouedec · 2026-10-16
- Databricks engineers land impactful PRs on day 2 with AI dev tools, vs week 3 before — Yuchenj_UW · 2026-10-16
- Perplexity open-sources SPACE, the sandbox platform powering its Computer agents — AravSrinivas · 2026-10-16