Modal Labs released state-of-the-art DFlash speculators for Qwen 3.5 and Qwen 3.6 models on Hugging Face, achieving 5-20% additional speedups and enabling Qwen 3.5 122B-A10B to run at over 1000 tok/s …
Modal and Z Lab released DFlash, a speculative decoding model for Qwen 3.5 397B-A17B, achieving over 4.3x throughput versus baseline and 1.5x versus MTP on HumanEval at concurrency 1. The model uses a…
AIWave has launched an API that provides access to over 50 Chinese AI models through a single OpenAI-compatible endpoint. The service supports models from DeepSeek, Zhipu, Qwen, and others, enabling d…
A developer fixed a broken fallback chain in their OpenClaw agent that was causing request timeouts during peak hours. The new chain includes seven entries: two local Ollama models, three OpenRouter f…
AgentHub has launched an open-source API gateway that provides unified access to over 100 AI models, including DeepSeek, Qwen, and Claude, at a flat rate of $0.14 per million tokens. This pricing is a…
Liquid AI released two new retrieval models, LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M, for fast multilingual search across 11 languages. The models, available on Hugging Face, are designed as dro…
Intrascope.app is building an AI workspace that gives teams centralized control over model access, usage, and costs. The platform supports BYOK and managed usage, allowing companies to manage multiple…
AIWave aggregates 50+ Chinese AI models behind a single OpenAI-compatible API endpoint, eliminating the need for multiple API keys, SDKs, and authentication schemes. Developers can switch between mode…
Researchers introduced the Argent Signaling Protocol (ASP) to mitigate semantic drift in multi-agent LLM systems by attaching structured quality signals to AI responses. In tests, ASP improved pass ra…
A developer with 8+ months of heavy agentic coding experience argues that low-cost LLMs like Qwen and Kimi handle 70-80% of software development work, while frontier models only earn their cost on the…
Qwen's coder outperforms Gemma 4 by 21 points on a benchmark, but the lead narrows significantly when both models run on a 16GB budget against a real repository, suggesting the bottleneck lies elsewhe…
Alibaba Cloud launched its fifth data center in Tokyo on June 18, 2024, just three months after opening its fourth, as part of an aggressive expansion in Japan. The new facility offers enterprise-grad…
China's 618 shopping festival showcases extensive AI integration by e-commerce platforms like Alibaba, but consumer adoption remains cautious, with shoppers embracing AI for price comparisons while re…
China is building a platform to debut new AI consumer products as part of its 'AI Plus' initiative, with over 70 domestic companies developing AI wearables like smart glasses. The initiative aims to e…
China's 618 shopping festival in 2026 hit a record 855.6 billion yuan in GMV, but weak consumer demand persisted as platforms stretched promotional windows. AI integration became central, with Alibaba…
A developer launched a public World Cup prediction arena pitting 12 AI models against each other, tracking 169 predictions across 21 settled matches. After 21 scoring entries, all models are tied with…
Researchers propose a structural pruning framework for Mixture-of-Experts (MoE) models that reformulates prune-ratio allocation as a channel-score coverage maximization problem, solved via attribution…
A developer building a Persian-language tutor using small Qwen-based agents shares five dataset design lessons learned from production failures, including avoiding tool-calling data where decoding can…
A software founder details how local Qwen models have provided real value for his business, paying for a GPU within months, but warns of infinite loops and hallucination risks, especially when quantiz…
A developer reports that the downloadsAllTime field for multiple public models on Hugging Face Hub is decreasing day-over-day, contrary to expectations for a cumulative count. The observed drops range…