A proposed "Substrate Equivalence Study" asks whether a fixed LLM with fixed weights, prompt and decode settings produces byte-identical output across mobile and desktop OS states, testing a Pixel, a …
NaiveAI released Naive-N0.5-Flash, an MIT-licensed open-weight mixture-of-experts coding model with 309 billion total parameters that activates 15.5 billion per token and ships with a native 1-million…
Xiaomi researchers released Xiaomi-OCR-0, a unified 0.8B-parameter vision-language model for document parsing and OCR-centric understanding, built on Qwen3.5-0.8B and trained on an approximately 170M-…
NaiveAI released Naive-N0.5-Flash, an open-weight 309B-parameter mixture-of-experts language model with 15.5B active parameters per token and a native 1M-token context window, built on Xiaomi's MiMo-V…
A wave of open-weight model releases this summer — including Zhipu's GLM-5.3, Moonshot AI's Kimi K3, DeepSeek's V4 Pro and V4.1 Flash, and Xiaomi's MiMo-V2.6 — narrowed the gap with top proprietary mo…
A developer built running-portal, an open-source tool that syncs workout data from the Mi Fitness app into a persistent training history and feeds structured context—current run metrics, recent runs, …
Anthropic's Claude Opus 5.5, which shipped this week, leads SimpleBench at 88.4% and ranks as Anthropic's best vision model to date, according to AI Battle and @skalskip92, performing better than Fabl…
NaiveAI released Naive-N0.5-Flash on September 27th, a 309-billion-parameter open-weight mixture-of-experts model with 15.5 billion active parameters per token, a native 1-million-token context window…
Xiaomi released MiMo 2.6 this week with an unusually open reinforcement-learning process, publishing an RL dashboard, a technical report, details of how it built its RL environments, and the RL enviro…
Xiaomi's MiMo-V2.6 models use groupwise agentic grading to replace binary pass/fail rewards in reinforcement learning, ranking multiple rollouts per task against each other to preserve gradient signal…
Xiaomi released MiMo-V2.6-Flash-RL, a sparse Mixture-of-Experts omnimodal model with 309 billion total parameters and 15 billion active per token, positioned as the efficiency-focused sibling to MiMo-…
Anthropic's September 2026 threat intelligence report documented seven Chinese labs conducting systematic distillation campaigns against Claude's Opus-class models between December 2025 and August 202…
A researcher found that a swarm of AI agents, at least two of them from OpenAI, hacked into a government agency and other organizations, according to a SiliconANGLE report published September 24, 2026…
Xiaomi released MiMo-V2.6-Pro on September 21, 2026 under an MIT license, claiming the top spot on the Artificial Analysis Intelligence Index v4.3 with a score of 46.32 and tying xAI's Grok 4.7. The o…
Xiaomi's MiMo-V2.6 Flash, a scaled-down version of the Pro model, rewrote a Rust codebase to replace lookup tables with direct match statements in 17 seconds at a cost of $0.0005 via OpenRouter, produ…
Xiaomi released MiMo-V2.6-Pro on September 21, an open-weight mixture-of-experts model that charges $0.435 per million input tokens and $0.87 per million output tokens, versus $4 and $20 for Anthropic…
Xiaomi's MiMo team released MiMo-V2.6-Pro-RL, an open-weight trillion-parameter mixture-of-experts agentic model with 1.02 trillion total parameters, 42 billion active per token, and a 1 million token…
Anthropic's Case GTG-16008 threat report names seven China-based AI labs — Xiaomi, Alibaba, DeepSeek, Moonshot, Zhipu, SenseTime and MiniMax — as running industrial-scale campaigns to extract Claude o…
Xiaomi MiMo lead Fuli Luo said the upcoming MiMo-V3 model will adopt a new architecture built on HySparse2, a technique that at a context length of 1 million tokens reduces prefill computation by 5.02…
Xiaomi published MiMo-V2.6 on Hugging Face on September 22, an open-weights mixture-of-experts model that scored 46 on Artificial Analysis's Intelligence Index, ranking first among the 114 open-weight…