Archive
GitHub 熱門專案、開源工具、框架、runtime 與重要 release。
推論系統
vLLM 0.28.0 can hierarchically offload the KV cache to disk and exposes an external secondary-storage manager, allowing long-context serving to move beyond the capacity limits of GPU and system memory alone. The release also advances Kimi-K3, DeepSeek V4, speculative decoding, and E/P/D disaggregation, but changes several defaults and dependency interfaces.
代理工具
The latest DeepSeek Harness release lets a primary agent select providers, models, reasoning effort, and output limits for subagents within its authorized scope. It also completes ACP interfaces for sessions, MCP, permissions, and cancellation. The release expands networking and external-model integrations, but the maintainers explicitly warn that its sandbox and approval mechanisms should not yet be considered secure isolation.
代理工具
Hillock combines factual triples, Hebbian association weights, and 10,000-dimensional hypervectors to determine on local hardware whether enough data exists to answer a question. Version 0.6 adds per-token MaxSim and multi-hop path encoding, but its current gating accuracy of about 56% shows that it remains a research prototype.
代理工具
Cogram has introduced a cloud-based modeling environment shared by humans and AI agents, using FreeCAD 1.1, the OpenCASCADE geometry kernel, and an MCP toolset to work with 3D models and technical drawings. Agents can already complete constrained, iterative modeling tasks, but engineers must still inspect complex models for geometry, dimensions, and solid constraints.
最新模型
ISTA DASLab has released three mixed-precision quantized versions of Qwen3.8-27B. GSQ learns the quantized values, while RCO selects the precision for each tensor under a fixed capacity budget. The smallest version is just 8.4GB and can be loaded directly by llama.cpp, Ollama, and LM Studio, although the near-lossless claim is still based on the team’s own evaluation.
GitHub Repo
OpenMAIC 1.0 expands its one-shot lesson generator into a course agent that can plan, revise, and resume work, with sessions stored through event streams and PostgreSQL. The new release provides 20 built-in skills and a vendor-neutral tool interface, but public deployments must still add identity isolation and scheduling for long-running media tasks.
最新模型
Z.ai has released the FP8 weights for GLM-5.3, ending a two-week safety buffer imposed because of the model’s increased offensive capabilities. The model can be deployed with frameworks such as vLLM and SGLang, but large MaaS providers are additionally subject to security review requirements.
代理工具
The new Channels feature can retain up to eight named sessions scoped by channel, chat, and sender, and reconnect to the original conversation after a session is closed. This is multi-task management, not parallel isolation: only one task can currently be selected at a time, and all tasks still operate in the same working directory.
推論系統
The new release adds `--usedirectio` for large GGUF weights and allows the relevant memory pages to be locked when a model is mapped with mmap. It also supports Qwen3.8-Flash-Next, Ling-3.0-flash, and custom JavaScript tools, but the project has not published loading-speed or memory benchmarks.
硬體
Sopro V2 Turbo distills a 500M-parameter teacher into a 120M-parameter model, then uses reflow to reduce acoustic solving from 32 steps to two. The model can produce its first audio chunk in about 300 milliseconds on an M3 CPU and supports in-browser inference, but it currently covers only four European languages and does not include training code.
GitHub Repo
Fizgig 5.0 updates all weights in MiniMax H3 and Krea 2 segment by segment, retaining gradients and optimizer states only for the current slice. Developer tests show peak VRAM usage as low as under 13GB, but system memory, training speed, and independent quality validation remain key constraints.
最新模型
Tencent has released Hy4 Preview’s Apache 2.0-licensed weights, an FP8 version, and deployment recipes for vLLM and SGLang. The model activates roughly 49B of its 770B parameters per token. It supports a 1-million-token context window, though evidence for its quality and self-optimization capabilities still comes primarily from Tencent’s internal evaluations.