Latest

Training code was the missing piece after the weights shipped, so this makes openPangu one of the few large MoE models reproducible end to end on non-NVIDIA silicon. Teams evaluating Ascend for large-scale training, or comparing Chinese open-source stacks against Qwen and DeepSeek, can now audit the full recipe rather than only run inference.

Primary sourcepandaily@[email protected] (Pandaily)Source

An MIT licence removes the commercial-use restrictions that still accompany many open-weight releases, so teams can fine-tune or ship it in products without negotiating terms. The hybrid SWA–DSA design without any full-attention layers is a notable architectural signal for anyone tracking how long-context efficiency is evolving beyond dense attention. Engineering teams evaluating open-weight alternatives for code generation and research pipelines should benchmark it against comparably sized MoE releases.

Primary sourcepandaily@[email protected] (Pandaily)Source

A food-delivery giant shipping a frontier-scale agent model shows how far Chinese consumer tech firms now push into foundation models, though the lack of published benchmarks makes real capability hard to judge. Developers building long-context or agentic workflows should track when independent evals and API access appear, since LongCat competes with open-weight MoE releases from Alibaba's Qwen and DeepSeek. Its 1M-token window targets codebase-scale agent tasks directly.

Primary sourcepandaily@[email protected] (Pandaily)Source

The release exposes full training pipelines tuned for Ascend hardware, making it a rare reference for teams that want to reproduce or fine-tune large models outside the NVIDIA ecosystem. Developers running Ascend clusters or evaluating Chinese domestic chips can use the repos as working baselines for pretraining and RL post-training. It also gives a concrete counterpoint to Meta's Llama-style open release playbook.

FAB measures whether models can combine images with instruction manuals, a skill that transfers to appliance repair and equipment inspection rather than simple document recall. Engineers building visual inspection or step-by-step guidance tools can use the 80% vs 28% gap as a marker of how quickly multimodal spatial reasoning has matured. Chinese models Alibaba Qwen and Google Gemini were also noted for characteristic error patterns in the study.

Meituan, better known for food delivery, is now shipping frontier-scale open models with 1M-token context, a signal that Chinese tech giants outside the usual AI labs are competing on agentic long-horizon workloads. Developers testing coding agents can plug it into Claude Code or Kilo Code directly and compare it against DeepSeek or Qwen on the same workflows. Availability via API and web makes hands-on trials straightforward.