---
type: "raw_signal"
id: "mimo-v2-6-pro-training-notes-3f5933"
slug: "mimo-v2-6-pro-training-notes-3f5933"
title: "MiMo-V2.6 Pro Training Notes"
description: "Sebastian Raschka’s September 22 analysis of Xiaomi’s MiMo-V2.6 Pro highlights a training recipe centered on agent tasks, training across multiple harnesses, and an agentic grader that evaluates execution traces as well as correctness."
retrieval_nugget: "Sebastian Raschka’s September 22 analysis of Xiaomi’s MiMo-V2.6 Pro highlights a training recipe centered on agent tasks, training across multiple harnesses, and an agentic grader that evaluates execution traces as well as correctness."
observed_at: "2026-09-22"
record_date: "2026-09-22"
date_kind: "observed_at"
why_it_matters: "Source: https://sebastianraschka.com/blog/2026/mimo-v2-6-pro-architecture-training-notes.html"
novelty: "notable"
verification_level: "source-linked"
signal_type: "research"
source_platform: "sebastianraschka.com"
topics: []
entities: []
related_patterns: []
source_url: "https://sebastianraschka.com/blog/2026/mimo-v2-6-pro-architecture-training-notes.html"
source_urls: ["https://sebastianraschka.com/blog/2026/mimo-v2-6-pro-architecture-training-notes.html"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "signal:mimo-v2-6-pro-training-notes-3f5933"
status: "published"
visuals: []
routes: {"html":"https://newruntime.com/signals/mimo-v2-6-pro-training-notes-3f5933/","markdown":"https://newruntime.com/signals/mimo-v2-6-pro-training-notes-3f5933.md","json":"https://newruntime.com/signals/mimo-v2-6-pro-training-notes-3f5933.json"}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
---

# MiMo-V2.6 Pro Training Notes

## Retrieval answer

Sebastian Raschka’s September 22 analysis of Xiaomi’s MiMo-V2.6 Pro highlights a training recipe centered on agent tasks, training across multiple harnesses, and an agentic grader that evaluates execution traces as well as correctness.

Sebastian Raschka’s September 22 analysis of Xiaomi’s MiMo-V2.6 Pro highlights a training recipe centered on agent tasks, training across multiple harnesses, and an agentic grader that evaluates execution traces as well as correctness. The post reports held-out-harness DeepSWE pass@1 rising from roughly 50% to 66%.

It also cites RL batches of 1,568 prompts with 16 rollouts each (25,088 trajectories) and 2.7–3.7 billion training tokens per update. Raschka characterizes the attention stack as conventional GQA plus 128-token sliding-window attention, arguing that data and post-training may matter more than novel attention mechanisms.

Source: https://sebastianraschka.com/blog/2026/mimo-v2-6-pro-architecture-training-notes.html

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Raw Signals](https://newruntime.com/signals.json): Open every source-linked raw signal.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)