追踪开放权重模型、开源 AI 工具和可复核的仓库更新。
test-llama-archs : generate dummy test vocab (#29084) Assisted-by: pi:llama.cpp/DeepSeek-V4-Flash-Vision-Exp **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11049/llama-
metal : support qwen4exp hc ops (#29000) Add support for the new DSV4 HC op variants used by qwen4exp: - hc_pre with per-element sigmoid gate (gated variant) - hc_post with identity mixing (comb == nullptr) Assisted-by: pi:llama.cpp/Qwen3.8-27B **Website:** -
## Bundled CLI Version This SDK release bundles CLI version: 0.24.1 Source: CLI built from source (same branch/ref as SDK) --- ## Bundled CLI Version This SDK release bundles CLI version: 0.23.3 Source: CLI built from source (same branch/ref as SDK) --- ## Bun
## What's Changed * fix(cli): scope the ACP permission queue to the session by @chiga0 in https://github.com/QwenLM/qwen-code/pull/11802 * feat(channels): add shared output modes with DingTalk support by @qqqys in https://github.com/QwenLM/qwen-code/pull/11822
## Highlights _See the complete change list below._ ## Breaking Changes - refactor(goal)!: stop emitting the active_goal stream event ([#12181](https://github.com/QwenLM/qwen-code/pull/12181)) by @qqqys ## Complete Change List ### Features - feat(workflows): l
## What's Changed `GET /api/show` now advertises each model's thinking controls and default: ```sh curl http://localhost:11434/api/show -d '{"model": "glm-5.3-flash:cloud"}' ``` ```json { "thinking": { "values": ["low", "high", "max"], "default": "max" } } ```
cuda : fix CUB argsort corruption caused by in-place keys (#28389) argsort_f32_i32_cuda_cub called the one-shot DeviceRadixSort::SortPairs API with d_keys_in == d_keys_out (temp_keys, temp_keys). CUB's internal double-buffer ping-pong requires distinct key buf
IT之家 9 月 19 日消息，阿里千问今日发布同声传译大模型 Qwen3.8-LiveTranslate ，模型以 Interleave 架构重构实时同传，准确度、流畅度、简洁度全面提升，字均延迟（LAAL）从 2.8 秒降至 2.3 秒。 官方表示，在支持 60 种语言的基础上，Qwen3.8-LiveTranslate 新增三项能力让同传更广泛的应用于真实场景中： 实时说话人分离，每句话归属清晰，音色复刻更稳定； 原文译文同帧同出，双语同屏； 长上下文消歧，联系前文读懂当下，人名术语翻译更精准。 Qwen3.
IT之家 9 月 19 日消息，据 21 世纪经济报道，9 月 19 日，华为计算战略与业务发展部部长朱照生表示，昇腾已经跨越了生态的拐点，完全有能力、有机会在 Agentic 时代构建 AI 新生态。 从社区来看，昇腾 CANN 开源社区月均活跃用户数超过 5200 人，自今年 6 月起成为国内活跃度最高的开源社区，社区非华为开发者数量超过华为开发者数量，目前日均新增合入代码超过 3 万行。 模型层面，基于昇腾完成预训练的大语言模型和多模态模型已超过 40 个，昇腾成为国内唯一支持预训练的 AI 算力平台。生态层
IT之家 9 月 19 日消息，科技媒体 Wccftech 昨日（9 月 18 日）发布博文，报道称 RPCS3 开发团队修复 NVIDIA GPU 驱动程序的一个 Bug，模拟游戏测试发现帧率可以提高 37%、显存占用降低 32%。 IT之家援引博文介绍，RPCS3 开发团队指出在 NVIDIA GPU 驱动程序中，存在影响模拟器性能的问题，会限制部分游戏的帧率，并在多数测试中推高显存占用。 贡献者 Yahfz 找到修复方案，可以变通缓解该驱动 Bug 影响，从而提高模拟器性能。开发团队在《GT 赛车 5》开放道
opencl: add support for bin kernel `flash_attn_f32_f16_bin` (#29046) * opencl: add `flash_attn_f32_f16_bin` * opencl: guarded prefill fa **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/rel
“IT早报”时间，大家好，现在是 2026 年 9 月 19 日星期六，今天的重要科技资讯有： 1. 12306 客服回应国庆高铁票涨价：日常票价有折扣，节假日有浮动是正常的 9 月 25 日，从重庆北到南京南的 D988 次列车票价是 446 元起，9 月 30 日同车次的票价则涨到了 520 元起。>> 查看详情 2. 水滴筹服务费上涨至 8% 引争议，回应称为了维持平台可持续运营、单季度亏损超 3600 万 患者家属申请提现 6,831 元，被收取 546.48 元的服务费以及 41.04 元的第三方支付通道
# Highlights *713 PRs from 237 contributors.* **New models in this release** (see the [cookbook](https://docs.sglang.io/cookbook) for all supported models): | Model | Type | PRs | Cookbook | |---|---|---|---| | GLM-5.3-Flash | Autoregressive | [#36507](https:/
hexagon: add ROLL op support (#29105) **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11045/llama-b11045-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI e
## What's Changed * docs(serve): record merged ACP boundary acceptance by @wenshao in https://github.com/QwenLM/qwen-code/pull/12024 * fix(ci): wait for the published export renderer before packaging the VSIX by @yiliang114 in https://github.com/QwenLM/qwen-co
hexagon: im2col update (#29103) * ggml-hexagon: accept 1D and padded IM2COL ops * ggml-hexagon: make pure-DDR IM2COL kernel is_2D-aware * ggml-hexagon: extend IM2COL DMA patch-embed fast path to 1D * ggml-hexagon: add blocked-staging general IM2COL DMA kernel
hexagon: HMX flash-attention head_dim padding (support DK=DV=72) (#26539) Allow HMX flash-attention to run with head_dim not a multiple of 64 (e.g. SigLIP head_dim=72), by operating on DK/DV rounded up to 64 with zero-filled tail lanes. **Website:** - **Attest
opencl: add bin kernel `kernel_gemm_noshuffle_q6_k_f32_32b_trans_ila_a8_bin` (#28678) * opencl: add A8 Q6_K non-MoE binary kernel * opencl: fix layout compatibility **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://githu
## What's Changed * docs(serve): record merged ACP boundary acceptance by @wenshao in https://github.com/QwenLM/qwen-code/pull/12024 * fix(ci): wait for the published export renderer before packaging the VSIX by @yiliang114 in https://github.com/QwenLM/qwen-co
Kimi K3 from Moonshot AI is now available on Amazon Bedrock, giving you a powerful new open-weight option for coding and knowledge work. It offers native vision, a 1-million-token context window, and explicit prompt caching to reduce latency and input costs.
Deploy production-ready Hugging Face models on Amazon SageMaker AI using six open-source agent skills. Point a coding agent at a model and get back a real-time endpoint with the right serving container, autoscaling, Amazon CloudWatch alarms, and a verified tea
ggml : check for allocation failures to prevent crashes (#28149) * ggml : check for allocation failures to prevent crashes * wording **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/release
Model-Saver: Write the SWA pattern, 15 more architectures roundtrip (#29042) * llama: read the SWA pattern as a period or a per-layer array Add llama_model_base::load_swa_pattern(), which reads sliding_window_pattern either as one flag per layer or as a period
ggml-webgpu: fix supports_op condition for GET_ROWS (#28978) * fix get_rows vec4 handling * Add src strides checking to vec4_aligned of get_rows and the new test case. **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://gi
ggml : handle graph buffer reservation failure (#26070) **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11036/llama-b11036-bin-macos-arm64.tar.gz) - macOS Apple Silicon
vulkan: add IQ3_S MMQ matmul kernels (#28822) * vulkan: add IQ3_S MMQ matmul kernels * Make block_a_to_shmem do 2-byte loads (110 bytes is divisible by 2) * Align the check, IQ3_S is also using K tile size **Website:** - **Attestations:** - **macOS/iOS:** - [m
### Summary of major features and improvements * #### More GenAI coverage and framework integrations to minimize code changes * New models supported: * On CPU: Gemma-3n * On CPU, GPU: Kokoro-82M, Qwen3-VL-4B with EAGLE-3, Qwen3-ASR, Muse Glimmer 30B, Qwen3.8 2
## What's Changed - Added first-run setup when running `ollama`, with options to sign in or continue locally. Setup completion is shared with the desktop app on macOS and Windows. - Added `ollama://apps` to open the desktop app’s Apps page directly on macOS an
## What's Changed * docs(serve): record merged ACP boundary acceptance by @wenshao in https://github.com/QwenLM/qwen-code/pull/12024 * fix(ci): wait for the published export renderer before packaging the VSIX by @yiliang114 in https://github.com/QwenLM/qwen-co
Mixture of experts (MoE) has become one of the defining architectural trends in large-scale AI model training. DeepSeek, Qwen, and Mixtral are examples of MoE...
