https://zlio.com/AI news, deduplicated.https://huggingface.co/blog/LiquidAI/lfm25-24b-a2b zlio-2868Tue, 13 Oct 2026 17:05:48 GMT LFM2.5-24B-A2B is an open-weight Mixture-of-Experts model from Liquid AI with 24B total and about 2B active parameters, designed to run fast, private agent tasks locally on a laptop. Weights are available on Hugging Face.https://www.reddit.com/r/LocalLLaMA/comments/1x5pjv7/the_ai_race_chinas_openweight_models_trail_us zlio-3671Wed, 14 Oct 2026 13:35:10 GMT A chart from whatmodel.sh, an open-source AI leaderboard, shows Chinese open-weight text models trail US closed frontier models by about seven months, based on 22,113 results across 987 benchmarks from 1,011 sources.https://news.google.com/rss/articles/CBMid0FVX3lxTFBJd3NDdURMc3BYVEU1VkpBemF6dFpuUTF4ZkZWNWNSeFY0ZHJtQzZMV1l6Sm00ZVpTbldfaEh6aXlYTUQybHRJNDQ3VnRtMUhNVHRqX3NNSEVGb0hxNU5DaUtEcVNnQU50V0xnekFPOEszQ0JEcW4w?oc=5 zlio-3836Wed, 14 Oct 2026 12:20:35 GMT FLock.io released an edge decision model built on Qwen aimed at enhanced data privacy, claiming it outperforms larger LLMs on selected tasks.https://www.reddit.com/r/LocalLLaMA/comments/1x5t1tm/nvidia_nemotron_bolt_35_74b_a9b_releases_oct_15th zlio-3883Wed, 14 Oct 2026 14:42:30 GMT A Reddit post says Nvidia's Nemotron Bolt 3.5 is a 74B-parameter, 9B-active MoE model quantized to 24GB using 2-bit quantization-aware training for speed.https://news.google.com/rss/articles/CBMiW0FVX3lxTE1rQ1h0SmhpbW9oVjd5QzFHNy12czVGR0lQcUIycWFoOUl6WDNIYUxXNHRraHhTYUFuaUdwbTZrX2k3RS1MWHdrLWd4ekZOT2diTS1UTkNUdGZXbzg?oc=5 zlio-3618Wed, 14 Oct 2026 12:08:00 GMT BigGo Finance argues Z.ai's GLM 5.3 shows post-training techniques, rather than raw compute, have become China's key competitive advantage in AI.https://www.marktechpost.com/2026/10/13/zai-glm-legal-harvey-lab-legal-ai-model zlio-2602Tue, 13 Oct 2026 08:19:54 GMT Z.ai's GLM-Legal, built on GLM-5.3-Flash, is a legal AI model scoring 19.2% all-pass on Harvey's Legal Agent Benchmark, nearly 3x its base, at $0.15 per test, by application.https://news.google.com/rss/articles/CBMiY0FVX3lxTFA4QWg0OHNvdTh0WHVGVG8tYkNOZjF2M3hRU25waXNtcGp2cFNnaFVSOVBCamNnVUtaOUtEQTE1Q0NySi1kWmw5VDJuYjYzeVAxdnVhMW9xaWZBVlg4VVU0Sk9YUQ?oc=5 zlio-3784Wed, 14 Oct 2026 14:01:04 GMT OrcaRouter contrasts a quietly published repo, HiRO-ACEv2, against the full launch of Qwen 3.8, weighing the two model efforts. The source is largely commentary without detailed benchmark numbers.https://x.com/Dheemanthredy/status/2110262830645510170 zlio-3170Wed, 14 Oct 2026 06:54:00 GMT An X post announces Maya-2, billed as the world's best Indic voice AI model with fully open weights, claiming it outperforms ElevenLabs v3, Cartesia Sonic 3.5 and Bulbul v3 across Indian languages; free to builders soon.https://www.reddit.com/r/LocalLLaMA/comments/1x5ak2n/someone_is_running_qwen_27b_at_20_toks_on_a zlio-2995Tue, 13 Oct 2026 22:21:00 GMT An enthusiast ran a 27 billion parameter Qwen model on a jailbroken PlayStation 5's GPU at about 20 tokens per second, demonstrating local LLM inference on console hardware.https://www.technologyreview.com/2026/10/13/1146366/open-models-have-real-benefits-do-they-outweigh-the-risks zlio-2646Tue, 13 Oct 2026 10:00:00 GMT MIT Technology Review Explains piece examines the benefits of open-weight models against risks, citing an Anthropic report arguing GLM-5.3 poses a major cyberattack risk.https://www.reddit.com/r/LocalLLaMA/comments/1x4sit6/based_on_advice_from_localllama_implemented_a_n2 zlio-2737Tue, 13 Oct 2026 09:57:21 GMT A hobbyist implemented an n+2 lookahead prediction for SSD-streaming Qwen Flash Next on a 64GB Mac Mini M5, fetching experts a layer ahead to speed token output.https://news.google.com/rss/articles/CBMihwFBVV95cUxQYTFpdjZnMjNiQnFOREhHblJ4WGpWUjRhaG1sS01VU2xpaDJsOHF3djI1ZVlFbEJiV1dZc1FaVy03QkVNalBNQV9xcnBDT3pvc1hiTDU1OUVRUS1fQ21JSC05VjhRRmdpSklTUldwaVZJMFdSMk9QdlE3VG1URWFOLVc2S1dXamc?oc=5 zlio-3032Tue, 13 Oct 2026 23:35:16 GMT OrcaRouter published first local timing tests running Qwen-Image-2.1-Turbo on a desktop via BaoCut 3.5, and compared it to Z-Image Turbo across two 8-step configurations.https://news.google.com/rss/articles/CBMif0FVX3lxTE02U1hiRFJqUjg5dkZaRDhZSExBbk8wNTNKb2REVmE3TUFueXRtOWpGeWx0d1pJY1Q3SktqeGYydmw4MkJ3eGFHM3p4cGg0T2d1emN6S0tsVnFqRmk0T3JHQ0czamVQRUpqdmprYnY2NmgyRWhYOGs3TmpJcEFMZVU?oc=5 zlio-3025Tue, 13 Oct 2026 23:27:20 GMT OrcaRouter compares two text-to-speech systems, Fish Audio Drama 3 Preview and Qwen Audio 3.0 TTS, on voice quality and expressiveness.https://www.marktechpost.com/2026/10/12/alibaba-qwen-introduces-qwengyre-elastic-reinforcement-learning-rl-for-xlong-horizon-agents zlio-2422Mon, 12 Oct 2026 21:41:50 GMT QwenGyre is an elastic online reinforcement learning framework from Alibaba's Qwen team for training extreme-long-horizon coding agents through unmodified black-box harnesses, targeting rollouts that run for hours.https://www.producthunt.com/products/kugelaudio zlio-2752Tue, 13 Oct 2026 13:51:15 GMT Kugel 3.5 is a text-to-speech model its makers claim outperforms other models on the market. It was shared on Product Hunt; no pricing or availability details were given.https://news.google.com/rss/articles/CBMiWkFVX3lxTFBnM3AtV29TbS1PeTZ1NDV1RmpGZE1OeTNldGtmT1hfRnNDV0JkUnk2V014UlJaSmN0cWIyS1V6UTBpY1lGMndxRTVBbEwySGFYMmJFVW5hdnNOQQ?oc=5 zlio-2749Tue, 13 Oct 2026 13:44:59 GMT A post examines Mistral Large 4's specifications, pricing and the absence of a stated licence. It is analysis of an existing model rather than a new release.https://news.google.com/rss/articles/CBMivAFBVV95cUxPUG5Kem5nVVNxU0NiOG81N21yRE0tblVTNDBmMlZYakdLS2QxQTh0M1dlOHdEd2xPQkpZdkdTWWpUYkFwSV9UQkdWd3FneDNON01VYzdvQzYzTFdTQXdnbEtyYXhkZTN3RWFyc3VWVWZJV25PaUtYNFZfUHdBTHUyQ1U2NkROQ0lnSUgwMXJEOFEyVHBENmRNUkE2UXVCOTd4RzA0UFEyNXF3VG9DNUdFdmFOX0tZdng5WnYyaQ?oc=5 zlio-2559Tue, 13 Oct 2026 05:15:12 GMT Brave New Coin reports DeepSeek and Xiaomi are leading a Chinese AI charge with 16 new models, challenging US tech giants.https://www.reddit.com/r/LocalLLaMA/comments/1x4uint/update_converting_dense_models_into zlio-2693Tue, 13 Oct 2026 11:50:03 GMT An independent developer continued work on converting dense LLMs into Mixture-of-Experts models, releasing Qwen2.5-1.5B-MoE-A0.9B along with GGUF versions for local use.https://eg2.haku.sh/ zlio-2681Tue, 13 Oct 2026 11:27:14 GMT A project repurposes Google's EmbeddingGemma 2 embedding model into a speech-to-text system, with the write-up and demo shared on Hacker News.https://news.google.com/rss/articles/CBMiZEFVX3lxTE1ybzFmTXV1Q1pOdlEyaHByaVlTUlk2bFpkVXJEdkN3NTk1MDRaZGlORUZjX3JJZUFjMy1pOV94bDhyOVhDamRHaVRzZEZiWGhGZmxHLVVSa01PRjJwMVQ2ZVBLdG4?oc=5 zlio-3361Tue, 13 Oct 2026 11:25:18 GMT OrcaRouter published a comparison of Qwen 3.7 and Qwen 3.8 aimed at helping developers decide which model generation to call for their workloads.https://news.google.com/rss/articles/CBMiWEFVX3lxTE1JSTIzbWxvNlljcGt0d3dTTDd3ck5fZU1CZFpqdFdNTndiWTNnWjI4SEluU29USGhxRnVOdUdseEdyWmpUWHZ4aFktcHhnQlppUFZNRXdTa00?oc=5 zlio-2981Tue, 13 Oct 2026 21:58:58 GMT An explainer article describing what DeepSeek is and how open-weight AI models work.https://www.reddit.com/r/LocalLLaMA/comments/1x4wcnn/is_everyone_in_here_buying_up_smart_tvs_now zlio-2735Tue, 13 Oct 2026 13:16:08 GMT A Reddit post jokes that Gemma4 26B A4B can be run with only 14 smart TVs. It is a community joke rather than a news item.