Latest releases
Nanbeige4.3, Flint and Raon
Fresh additions are ordered by verified publication date.
Current latest drops
Latest verified local releases are now tracked
The New page is hydrated from the same local-only catalogue selection as the homepage. If the dynamic cards do not appear, the newest entries remain available below.
Nov 4, 2026 · Workstation codingRipple 65B Preview48 GB memory recommendation · AD-Q4_K_M · Ripple-enabled llama.cpp
Oct 27, 2026 · Text and toolsLFM3 1.7B8 GB RAM estimate · Q4_K_M · LFM3-enabled llama.cpp
Oct 27, 2026 · VisionLFM3 VL 2B8 GB RAM estimate · Q4_0 · LFM3-enabled llama.cpp
Oct 24, 2026 · Coding and toolsNanbeige4.3 33B32 GB RAM estimate · Q4_K_M · Start with short context
Oct 19, 2026 · Local agentsSpark-X2.6 4B8 GB RAM estimate · Q4_K_M · Start with short context
Oct 16, 2026 · Recurrent reasoningNVIDIA Dory 1.0 4B16 GB RAM estimate · Q5_K_M · Custom runtime required
Oct 12, 2026 · Agent MoELFM2.5-24B-A2B24 GB RAM · Q4_K_M GGUF · 131K context
Also in the current freshness set: MiniCPM-V 4.7 35B-A3B, Mellum2.1 12B-A2.5B Thinking, MiniCPM-V 4.7 (1B), MiMo-V2.6-Flash-RL, Xing4.0-29B-A4B, Bonsai 2 27B, LLM-jp-4.1 32B-A3B Thinking, Needle 3, LLM-jp-4.1 8B Thinking.
Catalogue
256 verified local LLMs
Downloadable models with a practical path for running them locally.
Speech catalogue
120 local voice tools
Browse local text-to-speech, speech-to-text and audio workflows.
Try it now · Browser playground
Open the decision lab →Meet the small models in Labs.
Run Qwen3, MiniCPM5 and Qwen3.5 on your device. Free chat, games and 18 decision situations. No account or API key.