#On-device

AILiquid AI announced its on-device Liquid Context layer is now optimized for Snapdragon processors using the Qualcomm Hexagon NPU, letting edge agents learn user routines and share context across devices. Separately, Liquid AI released LFM2-1.2B-Longevity and LFM2-2.6B-Longevity, which the company says often match or outperform much larger frontier LLMs on longevity prediction tasks.

AILiquid AI released an experimental DSpark draft model for its LFM2.5-VL-3B vision-language model, delivering decoding throughput gains of up to 2.66× on GPUs and 3.13× on edge devices. The drafter adds about 280M parameters, an 8.9% increase in the deployed model's parameter count, and is available on Hugging Face with support in llama.cpp, SGLang, and MLX-VLM.

AIGoogle announces support for local AI models in the Antigravity SDK, per a linked developer blog post. The post itself offers no further details, so specific features, supported models, or limits cannot be confirmed from this source.

AIGoogle announced that the Antigravity SDK supports local agent workflows, with initial support for Gemma 4 26B A4B through Google AI Edge's LiteRT. The post includes Python setup steps and says a recommended machine has more than 24GB VRAM or unified memory. It also describes a hybrid pattern in which a cloud Gemini 3.8 Flash planner hands work to local Gemma 4 26B models, with 97.2% of tokens in one recorded run staying local.

Why it matters: The source shows how to run an agent with a local Gemma 4 26B model using LiteRT, plus a hybrid cloud-planner pattern that keeps most tokens on-device.

AIHugging Face's transformers library can now load the same GGUF checkpoints used by llama.cpp, with fast local inference on Mac powered by ggml's Metal kernels. The author, Georgi Gerganov, says the work brings those kernels into the transformers ecosystem to increase compatibility and performance.

AIA developer connected a Raspberry Pi Zero W2 with a webcam and ambient microphone to MiniCPM-o 4.5, an open-source model running locally on a PC. The setup streams audio and video in real time with full-duplex interaction and no cloud APIs.

AIOpenMed reports that MiniCPM5-2B, run on a local Mac, correctly kept naproxen in medication history rather than the current-medication export after a newer note said it was stopped. Every graph connection in the run links back to its source, using fictional clinical notes.

AIA developer's Augury plant identification model, built on an OpenBMB model, raised photo top-1 accuracy from 71.8% to 80.2% by merging duplicate species keys and adding PCA whitening. The next steps are reaching 90%+ accuracy and building a phone GUI so farmers can use it on-device.

AIA commenter praises a practical setup where a small local model runs a log-debugging pipeline entirely on CPU through OpenClaw and Ollama. They argue that workflow design can make a small model far more useful than its raw size suggests.

AIOpenBMB's 2B-parameter MiniCPM5 model runs as a local news desk on an older i5-9400F PC with 16GB RAM and no cloud API. The developer built a system that collects official sources hourly and sends a 24-hour Telegram recap with a lead story and links.

AILM Studio announced that Qwen3.8-27B runs at up to 144 tokens per second on an M5 Max MacBook Pro through its partnership with Inco Splash. The post claims up to 3× the decode speed of Ollama, 2× oMLX, and almost 4× when an agent fans out into sub-agents. Inco Splash is described as an open-source inference engine built for the model and Apple silicon, available through the linked LM Studio blog.

AIKDE's Akademy conference in Graz, Austria, opens on September 19, where contributors Eva Brucherseifer and Jan Muehlig will present a talk proposing an "AI-native" KDE desktop built on a personal, encrypted "Kadai" kernel. The proposal's middle section is expected to divide attendees, while the project marks its 30th anniversary, with KDE 1.0 released in July 1998.

AIOpenBMB's MiniCPM5-2B is a 2.5B-parameter model with native 128K context, offering hybrid Think and No-Think modes in one checkpoint. Users can download it from Hugging Face and run it fully offline on-device, as RunAnywhere demonstrated. In a demo, the model first called a puzzle impossible, then corrected itself and wrote a working verifier.

AIOpenBMB has released MiniCPM5-2B, a dense 2B Transformer built for on-device and resource-constrained deployment, with an average score of 53.9 in its comparison set. The release also opens the UltraData datasets behind it, including UltraX, UltraData-Code, UltraData-SFT-Agent-2609 and UltraData-RL-2609, and includes GGUF, MLX, GPTQ and DSpark variants for common runtimes.

Why it matters: The release pairs a 2B model with open training datasets and reports per-benchmark comparisons against named same-size and larger models, letting readers check the claims directly.

AIA Berkeley Sky Lab researcher says stronger open models and new inference systems will make powerful local AI practical. The linked post reports Qwen3.8-Flash-Next running at 68.3 tok/s on a single RTX 5090 using an NVFP4 checkpoint, with 63GB host RAM and a 51GB n-gram table stored on NVMe at about 0.5% throughput cost.

AIUnsloth Desktop now runs GLM-5.3-Flash GGUFs out of the box with faster inference, enabling MTP and faster long-context decoding. The quoted Unsloth post reports local GGUF inference 1.6–3.4× faster with optimized decoding and multi-token prediction, and 3-bit runs on 128GB setups.

AIHermes now includes a local backend that runs Unsloth UD-Q4_K_XL and UD-Q4_K_M quants for DeepSeek-V4-Flash, Qwen3.8-27B, Qwen3.6-35B-A3B, and Qwen3.8-Flash-Next. The post says these models can be run locally in one click, with Unsloth GGUFs supported.

AIMirai is releasing speculative decoding in its Uzu inference engine, starting with Qwen3.6-27B. The quoted post reports 105 output tokens per second on an Apple M5 Max with 128 GB of unified memory, 2.9× faster than the fastest MLX speculative-decoding implementation Mirai benchmarked. The stack combines DFlash with Mirai's Weaver model, tree-based speculative decoding, Mirai quantization, and Metal kernels for Apple silicon.

AILiquid AI and Artificial Analysis released Pipette, an open-source benchmark platform for foundation models on edge devices, covering over 1,000 configurations across 30+ models. It measures five on-device metrics, including throughput, latency, context scaling, and memory use, on macOS, Windows, iOS, and Android. Liquid AI also said its updated LFM2.5 Q4_0 checkpoints, trained with Quantization-Aware Distillation, retain roughly 97% of BF16 baseline performance and suffer 73.4% less quality loss than standard post-training Q4_0 quantization.

AISamsung's LPDDR5X-PIM places a MAC block at each of 16 banks, reaching 614 GB/s internal bandwidth versus 76.8 GB/s for regular accesses. Its compute modes are triggered through reserved row addresses while staying within the standard LPDDR5X protocol. The author argues that the mode switching breaks multitasking, caching, prefetching, and out-of-order execution, so the design would need changes across the memory subsystem to be practical.

AIOpenBMB released MiniCPM5-2B-SFT, an SFT-only BF16 checkpoint taken before RL and OPD, within its MiniCPM5-2B series. The model is a 2B dense Transformer built for on-device and local deployment, with 131,072-token context and the same training recipe as the final release.

Why it matters: The source gives concrete benchmark averages against same-size and larger models, plus released training data and multiple deployment formats, useful for judging a compact on-device model.

AITinker announces Coco, a local assistant that proactively offers help when users are unsure what to ask, with a voice interface powered by Inkling on Tinker. According to the background post from @EchoShao8899, the voice trigger is a "Hi Coco" feature, and the part was open-sourced first.

AISky Lab's FreeToken runs official checkpoints of large models such as Qwen3.6 35B on an 8GB RTX 4060 laptop at 39 tokens per second. The same approach reportedly serves DeepSeek-V4-Flash 284B at 22-25 tokens per second on an RTX 5090 desktop, and GLM-5.2 753B at 15 tokens per second on an RTX PRO 6000 workstation.