AI Briefing#long-context
The latest AI and developer news about #long-context, with the original source and a short summary.
The latest AI and developer news about #long-context, with the original source and a short summary.
Meta has released Muse Spark 1.3, an AI model with significantly improved 1M context retrieval performance and coding efficiency.
A new study found that Sliding Window Attention outperforms linear attention by 2 to 10 times on long-context reasoning benchmarks, raising questions about the need for post-processing.
dots-studio has released a 280B MoE multimodal model supporting a 512K context window.
Chunked KL loss reduced Knowledge Distillation memory usage from approximately 85GB to 5GB for 32K contexts.
Zero-Mem proposes a structure for retrieving and managing agent memory without LLM calls.
LongCat-Flash-Lite-Sparse weights have been released, supporting up to 1M tokens of context.
Poolside updated the FP8 and NVFP4 checkpoints for Laguna S 2.1.
A new benchmark called HANDBOOK.md has been released to evaluate the ability of AI agents to perform complex tasks while complying with long policy documents.
A new benchmark called Handbook.md has been released to evaluate whether AI agents perform tasks while complying with lengthy policy documents.
Liquid AI has released two LFM2.5 encoder models capable of fast long-context processing even in CPU environments.
inclusionAI released LLaDA2.2-flash, an agent-oriented diffusion language model supporting 128K context and Levenshtein Editing.
Upstage has released Solar Open 2, a 250B-scale open-weight MoE model that supports agentic workflows and 1M token context.
MoBA, a technology that applies the MoE principle to attention to maximize the efficiency of long-context processing, has been announced.
A German research consortium has released Soofi S, a 30B model that surpasses OLMo 3 and Apertus on fully open-source LLM benchmarks.
LG AI Research has released 3 EXAONE 3.5 open-source models that maximize performance and efficiency.
LG AI Research has released 3 EXAONE 3.5 models as open source, capturing both performance and cost-efficiency.
Moonshot AI's Kimi K3 has become available on Vercel AI Gateway.
MiniMax M3 uses Sparse Attention technology to dramatically reduce cost and computation even with long contexts, enabling the implementation of long-horizon agents.
Deep Agents introduced RLM and dynamic subagents to improve performance in handling large-scale context.
The PageStorm research model, designed for book-scale creative writing, has been released.
The Qwen team announced HydraHead, an architecture that combines FA and LA at the attention-head level to improve long-context processing performance.
The open-source model GLM-5.2, featuring a 1M token context and improved coding capabilities, has been released.
MiniMax has announced MSA, a new sparse attention technology that maximizes computational efficiency for large-scale context processing.
silx-ai's Quasar-Preview model, supporting a long context length of 5 million tokens, has been released on Hugging Face.