MiMo V2.6 Flash MOPD with thinking off is very bad. on LiveCodeBench: accuracy: 55.5 avg tokens per sample: 6k for comparison Qwen3.8 Flash Next with thinking off: accuracy: 75.8 avg tokens per sample: 2.8k (btw, Qwen3.8 Flash Next is one of the best models with thinking disabled, but still completely useless mode for agentic tasks)

MiMo-V2.6-Pro is available today as an on-demand deployment. → Highest-scoring open-weight model on the AA Intelligence Index (46) → 1.02T total / 42B active MoE → 1M context, multi-modal, with MIT license Spin up an endpoint and tell us what you’re building →

MiMo-V2.6 just dropped, and Xiaomi’s Fuli Luo is already teasing the next architecture: MiMo-V3🔥 The core of it, HySparse2, targets three bottlenecks in agentic inference: prefill cost, KV cache size, and long-context retrieval. Compared with MiMo-V2.6’s Hybrid SWA architecture at 1M tokens: • 5.02× lower prefill FLOPs • 4.5× smaller KV Cache • Better MRCRv2 and RULER-v2 scores, plus lower AgentPPL and LongPPL The HySparse2 paper comes from Xiaomi’s LLM-Core team, with Fuli Luo as corresponding author and Team Lead. Interestingly, its references also include DeepSeek-V2, V3.2, V4, and V4.1-Flash, along with OpenAI’s GPT-4.1 evaluation work and the gpt-oss model card.

MiMo 2.6 Pro vs Flash👇 Pro (left): 10M tokens, $1.41, 4,927s Flash (right): 6.9M tokens, $0.15, 4,090s Pro costs nearly 10× more, but the night scene, lighting, and overall atmosphere are noticeably richer.

MiMo-V2.6-Flash is available for free for a week on OpenCode! if your looking for a way to test this model out... you can now for free which is pretty amazing let me know what yall cook with it

MiMo-V2.6: The Hard Road to Scaling Up RL MiMo-V2.6 is very likely one of the largest single RL runs, by compute, that any open-source model team has undertaken to date. In an era when compute is brutally scarce, we still chose to dedicate a team of several dozen people to one goal over an extended period: scaling up RL. That takes more than research conviction. It takes a vision for AGI, respect for the unknown, and the nerve to walk straight into the hardest problems. The result is a model whose potential was built through mid-training and unlocked through heavy RL. Today, it is the number one open-source model. I strongly recommend reading the technical report. I believe it will become one of those papers that Agent RL practitioners keep reopening and discovering something new in each time. In my view, the research innovations and engineering challenges behind it surpass those of DeepSeek R1, which I was partly involved in. Some will ask: why MixRL instead of MOPD? First, they are not competing choices. We ran MixRL on verifiable tasks of moderate difficulty, including code and related agentic tasks, and found that the resulting models generalize remarkably well. Second, tasks that are difficult to verify, extremely long-horizon, or simply too challenging to include in a joint RL run are trained separately. Including them would substantially reduce rollout efficiency or introduce significant rollout staleness. We then merge the resulting capabilities through MOPD. Games, 3D tasks, and tasks with subjective evaluation signals all fall into this category. There is also a third, slightly cheeky answer. Our team is flat enough and free enough of organizational silos that MixRL simply is not difficult for us. More importantly, everyone enjoys working this way. People from different domains come together every day, driven by the pursuit of AGI and intelligence that can continuously improve itself, to confront and resolve the RL bottlenecks in each field. I will always remember the RL daily update meetings from this period. They were intense and dense, with intelligence emerging in real time. To help the open-source community focus on solving real Agentic RL problems, we have released a Qwen model distilled from MiMo RL trajectories as a stronger starting point for RL, along with 7K diverse environments and a complete RL training framework. We hope these resources will help move Agentic RL research forward. MiMo-V2.6 is only the beginning. In an era when intelligence is easy to replicate, we still choose the hard road toward self-improvement and AGI. Much of what lies ahead remains unknown. But we are willing to keep investing the time, compute, and passion required to take on one hard problem after another and work each of them all the way through, until intelligence crosses into a new regime.