Mach-Mind-4-Flash Technical Report
2026-07-10 • Machine Learning
Machine LearningComputation and Language
AI summaryⓘ
The authors developed Mach-Mind-4-Flash, a large AI model that uses a special approach to activate fewer parameters but still performs as well as much bigger models. They improved training efficiency by combining different expert models and using a new training technique that prevents performance issues common in complex reinforcement learning. Their methods also compress reasoning steps without losing much accuracy. The resulting model performs very well across various tests while being computationally cheaper than much larger models.
Mixture-of-Experts (MoE)Reinforcement Learning (RL)Multi-Teacher On-Policy Distillation (MOPD)Policy OptimizationToken EfficiencyParameter ActivationAgentic ModelTraining AccelerationReasoning ChainsModel Compression
Authors
Foundation Model Team
Abstract
We present Mach-Mind-4-Flash, a 35B-parameter Mixture-of-Experts (MoE) agentic model with 3B activated parameters. Through post-training optimization alone without scaling pre-training compute, the model achieves performance on par with or surpassing that of 100B-parameter-class models. By introducing scalable agentic interaction environments for large-scale reinforcement learning, the model attains significant performance gains on real-world application tasks. Our pipeline comprises three stages: (1) a unified RL/OPD training infrastructure with dynamic multi-teacher scheduling and operator-level acceleration, delivering 17\% end-to-end training speedup; (2) multiple domain-specific RL experts trained in parallel across Reasoning, General, and Agent tracks, then fused into a single generalist via Multi-Teacher On-Policy Distillation (MOPD) -- a routed reverse-KL objective that eliminates the see-saw degradation of mixed-reward RL; (3) Hybrid Median-length Policy Optimization (HMPO), a single-stage token-efficiency method that compresses reasoning chains by 19--46\% with $\le$0.7 percentage-point accuracy loss. Mach-Mind-4-Flash scores 92.70 on AIME'26, 82.82 on IFBench, 80.74 on Behavioral-SafetyBench, 75.80 on BFCL-v4, 72.31 on BrowseComp-zh, and 84.20 on ClawBench -- leading or matching models with 10--30$\times$ its activated size at a fraction of the inference cost.