llama.cpp improves selected GPU inference paths for low bit models
A merged llama.cpp optimization brings hardware specific gains. Here is what the measurements show and how to check whether your setup benefits.
Qwen
Qwen Code 0.25.1 adds side tasks and clearer approvals
Qwen Code 0.25.1 expands side conversations and approval previews. A later background agent change has a separate availability boundary.
Janas adds Windows support for local AI models larger than RAM
The experimental runtime now has Windows and Linux packages. NVMe storage and available memory still shape what it can run and how well it performs.
Boole 1.7B compresses Qwen for local inference
Precision Numerics releases binary weights and local runtimes with developer reported quality tradeoffs.
Infernix Windows release fixes cache exhaustion
What to check before running the new local Qwen engine.
QwenCloud adds Qwen Image 2.1 Pro API access
The October 2 hosted API addition supports transparent images and editing with up to ten references.
NVIDIA adds DGX Spark 64GB from $4,999 for October 23
The partner configuration keeps GB10 while cluster setup and model deployment remain separate steps.
Qwen 4 code trail points to Huawei Ascend support
A September 30 SGLang draft targets Huawei Ascend attention support around the Qwen 4 architecture preview. Complete model serving still needs validation.
Alibaba introduces Qwen 3.8 Max with API access
Alibaba launched Qwen 3.8 Max with global API access and plans for model weights. The launch announcement did not verify the pricing quoted in the earlier article.
