After Xiaomi MiMo-V2.5's weights went public, I checked whether it runs on Mac/ROCm or on cloud GPU (RunPod/GCE). It's still rough on local hardware, but RunPod's 4x H200 runs it for ~$14/hr and GCE Spot H100 brings it down to ~$1.6/hr.
SB Intuitions released sarashina2.2-tts, an LLM-based TTS model focused on Japanese. It clones speaker voice and style from short reference audio without fine-tuning, and handles Japanese-English code-switching.
Inclusion AI released LLaDA2.0-Uni. A 16B MoE diffusion LLM that handles image understanding, 1024px image generation, image editing, and interleaved text-image generation in a single model.
An LLM safety monitor's evaluator can be tricked into clearing dangerous sessions when the attacker plants fake analysis text in the monitored conversation. Experimental results, defense limits, and structural separation points.
A look at WUPHF's LLM wiki that trended on Show HN. Using Markdown as the source of truth with Git history, SQLite, Bleve, and lint for shared agent memory is a different bet from RAG or managed memory.
OpenAI shipped GPT-5.5 and GPT-5.5 Pro on the API. A practical rundown of the 1M+ context, the new reasoning.effort default, image input behavior, prompt caching, and pricing.
Hands-on running inclusionAI Ling-flash-2.0 (100B / 6.1B active, MXFP4 quant, 54.7GB) on SwiftLM via mlx-swift-lm on an M1 Max 64GB. Covers bailing_moe + MXFP4 support check in mlx-swift, the startup surprise, and what --stream-experts actually saves.
A hands-on build and run of the Swift-based LLM inference server SwiftLM on an M1 Max 64GB. Covers Qwen3.6-35B-A3B and Qwen3.5-122B-A10B, with the same BST, BBS, and persona tests used in the existing Ollama and MLX-lm write-ups.
TRACER, a recent arXiv paper, takes the input/output logs of an LLM classification endpoint and reuses them as training data, then swaps in a lightweight surrogate only on regions that pass a parity gate to cut inference cost. The surrogate absorbs 83–100% of traffic on a 77-class intent dataset and 100% on a 150-class one, while correctly refusing to deploy on an NLI task — that refusal behavior is the interesting part.
Japan's Digital Agency released parts of Gennai, the generative AI platform it runs for central-government staff, on GitHub under MIT / CC BY 4.0. The web app and cloud-specific AI templates for AWS, Azure, and Google Cloud are bundled together so local governments and private companies can redeploy the same stack.
DeepSeek V4 Preview ships V4-Pro (1.6T/49B active) and V4-Flash (284B/13B active) as open weights under MIT, both with 1M context. CSA+HCA hybrid attention, mHC, and the Muon optimizer cut per-token FLOPs at 1M tokens to 27% of V3.2. Day-one API and chat.deepseek.com mode switch covered.
Two open-weight Chinese MoEs landed within 24 hours: Ant Ling-2.6-flash (104B/7.4B active, 7x token-efficiency claim) and Tencent Hy3-preview (295B/21B active, frontier-tier open weights). Specs, licenses, and how they line up against DeepSeek-V3 and GLM-4.5.
