Xiaomi MiMo-V2.6 details its largest RL scaling run: $2.6M, 1M-token contexts
KyeGomezB · x · 2026-09-22
Xiaomi's MiMo team released a technical report on MiMo-V2.6, its largest RL scaling run to date, with open-sourced training dynamics, RL environments and framework.
- Cost: $2.6M for Pro, $0.9M for Flash; 44% spent on rollouts, 41-44% on training, 14% on grading.
- Scale: each RL step consumes 1,568 prompts × 16 rollouts, producing 25K trajectories and 2.7-3.7B tokens at contexts up to 1M tokens.
- Techniques: groupwise agentic grading — comparing successful trajectories and shifting reward toward shorter, higher-quality solutions; frozen MoE router plus multi-layer defenses against reward hacking.
- Domains: coding, general workflows, visual tasks, cybersecurity.
- Results: DeepSWE improved from 58.4 → 72.6 (Pro) and 48.7 → 65.7 (Flash).
More from Models
- How to cut Fable 5.2 token costs by 90%: Arc routing plus a lean CLAUDE.md — EXM7777 · 2026-11-12
- Partial RuneBench results: Inkling 2 Small scores 5.314, far above original — mariofilhoml · 2026-11-12
- Grok Bot 1.0 Announced: Shareable Bots With Upgraded Collaboration — SpaceXSI · 2026-11-12
- Polymarket odds put Chinese AI lab at #1 by year-end at just 6% — Polymarket · 2026-11-12
- FutureHouse's deslopper trains Kosmos to write indistinguishably from human experts — proteinrosh · 2026-11-12
- Developers say Gemini 4 Carbon is notably better at favicon design and micro-interactions — AI_Andrew · 2026-11-12