315M-param self-play RL bot Pluto crushes GPT-6 Astra 0-1000 in StarCraft: Brood War
i_dg23 · x · 2026-10-05
The LLM-plays-StarCraft challenge ended in a rout: GPT-6 "Astra" went 0-1000 against Pluto, the top human-made Brood War bot.
- Pluto is a 315M-parameter hybrid transformer/RNN trained via self-play RL like AlphaStar — essentially a solo project by tscmoo (Vegard Mella, Meta)
- It held a 98% winrate against all other bots at IEEE CoG 2026, has beaten multiple pro BW players, and a bootleg version hit rank 5 on the global ladder
- To make BW an RL env, tscmoo re-implemented the core game engine in 30K+ lines of C++ (OpenBW), mostly written in 2016 pre-LLM assistance
- Pluto is exploitable, the author notes — maybe GPT-7 should train its own deep RL model for revenge
Related event: Self-play RL bot Pluto crushes GPT-6 Astra 1000-0 in StarCraft(3 posts)→
More from Research
- Meta's NAVA-WAM pretrains robot action policies directly from action-free videos — meta · 2026-10-05
- gamfit: open-source Rust engine fits GAMs from a formula with REML-chosen smoothing — Sauers_ · 2026-10-05
- Fine-tuned Llama 3.1 8B for medical decisions hits 84% per-field accuracy, only 30-34% perfect rows — Forsaken_Cut8542 · 2026-10-05
- SJTU's LIFT adds force sensing to VLAs with zero force-labeled pretraining data — jiqizhixin · 2026-10-05
- LOOM stabilizes looped MoEs at 9-12 loops, beating standard MoE at iso-FLOP — SonglinYang4 · 2026-10-05
- Paper accepted at ACML 2026, but author can't afford to present it — Jealous_Key_4030 · 2026-10-05