Topic: LLMs Related concepts: Context Engineering RAG Frequently mentioned with: Hugging Face NVIDIA Kimi K3.1 Pro OpenAI Apple Recent briefs: 2026-10-26 2026-10-25 2026-10-23 2026-10-22 2026-10-20
Topic: LLMs Related: Context Engineering RAG
Recent Updates
- 2026-10-09: Galahad-KV Enables Fast and Cheap KV State Reuse Across a 50M-Token Stream via NVMe Offloading (breakingnewsofficial) · arxiv.org — NVIDIA · Gemma 4 12B · Gemma 4 31B
- 2026-10-09: ByteDance Seed Discovers Phase Sensitivity Causing Periodic Retrieval Fluctuations in DeepSeek Blocked KV Cache Compression (breakingnewsofficial) · qbitai.com — ByteDance · DeepSeek · QbitAI RunningHub · MiniMax Wen Le · DeepSeek V4 DeepSeek-V4-Flash-Base DeepSeek-V4-Pro-Base · DeepSeek-V4-Flash-0731 DeepSeek-V4-Pro-0813 DeepSeek-V4.1-Flash-0910 · Qwen3-0.6B MiniMax H3
- 2026-10-12: KV-Repurpose: An Open-Source Testbed for Non-Prefix KV-Cache Reuse in LLM Serving (breakingnewsofficial) · arxiv.org
- 2026-10-12: LongHarness: Stateful Evidence Management for Agentic Reasoning over 100M-Token Corpora (breakingnewsofficial) · arxiv.org — GLM-5.3-Flash · Qwen3.8-Flash
- 2026-10-13: Liquid AI Releases LFM2.5-24B-A2B Reasoning Agent Model for On-Device Inference (breakingnewsofficial) · huggingface.co — Liquid AI · Hugging Face · Apple · NVIDIA LFM2.5-24B-A2B LFM2-24B-A2B · Qwen3.6-35B-A3B Gemma-4-26B-A4B-it · Nemotron 3.5 Lightning
- 2026-10-15: LFM2.5-24B-A2B: Open-Weight MoE Reasoning Model for Fast On-Device Agents (breakingnewsofficial) · liquid.ai — Hugging Face · Apple · NVIDIA · GitHub · PyPI LFM2.5-24B-A2B LFM2.5 230M LFM2.5-24B-A2B-Base LFM2-24B-A2B · Qwen3.6-35B-A3B
- 2026-10-15: Apertus 1.5 Release: Multimodal Understanding and Extended Context Window (breakingnewsofficial) · huggingface.co — Hugging Face Swiss AI Initiative Apertus 1.5 Apertus 1.0
- 2026-10-15: Meta Releases Muse Spark Open 816B Parameter Open-Weight Model (breakingnewsofficial) · research.meta.ai — Meta · Apple · Baseten · Fireworks AI · Groq · Modal · Nebius · OpenRouter · Together AI · Amazon · Google Cloud · Hugging Face · Microsoft · AMD · Intel · NVIDIA · Muse Spark Open
- 2026-10-16: From Long Context to Selective Memory: A Survey of Memory-Centric Compression for Large Language Models (breakingnewsofficial) · arxiv.org
- 2026-10-16: Hugging Face Introduces Beam, an Open 501B Sparse MoE Model for Coding and Agents (breakingnewsofficial) · huggingface.co — Hugging Face · Reflection Beam DFlash2 · DeepSeek-V3
- 2026-10-16: Local and Distributed Large LLM Inference with MLX and RDMA on Mac Studio (breakingnewsofficial) · machinelearning.apple.com — Apple · Hacker News · Muse Spark Open Mistral Medium 3.5 · Qwen 3.8
- 2026-10-17: NVIDIA Shares Open Midtraining Recipe for Agentic Coding Models (breakingnewsofficial) · arxiv.org — DAIR.AI · NVIDIA · Nemotron 3.5 Lightning Nemotron 3.5 Super
- 2026-10-19: Artificial Analysis Evaluates DeepSeek V4.1 Flash 1017 (Max) (breakingnewsofficial) · artificialanalysis.ai — DeepSeek · Artificial Analysis DeepSeek V4.1 Flash 1017 (Max)
- 2026-10-22: The Bottleneck Is Retrieval, Not Reading: Diagnosing Evidence-Based Generation with RAG for Long Context Question Answering (breakingnewsofficial) · github.com
- 2026-10-22: Don’t Stop Long-Context Continual Pre-Training Too Early: An Empirical Scaling Law (breakingnewsofficial) · arxiv.org
- 2026-10-23: Position Bias Across Long-Context Model Architectures: How Evidence Position and Prompt Structure Affect Performance (breakingnewsofficial) · arxiv.org
- 2026-10-23: Demystifying True LLM Call Costs: Formulas, Prompt Caching, Reasoning Tokens, and Long Context (breakingnewsofficial) · habr.com — Habr
- 2026-10-23: Moonshot AI Announces Kimi K3.1 Pro Open-Weight Model Series (breakingnewsofficial) · kimi.ai — Hacker News · Kimi · Unity · Epic Games · Kimi K3.1 Pro Kimi K3.1 Pro Highspeed · Kimi K3 · GPT-6 Astra · Claude Opus 5.5
- 2026-10-23: Kimi Code Releases K3.1 Pro and K3.1 Pro Highspeed Models (breakingnewsofficial) · kimi.com — OpenAI · Anthropic K3.1 Pro · K3.1 Pro Highspeed · K3 K2.8 Preview K2.7 Code Highspeed
- 2026-10-23: Kimi.ai Launches K3.1 Pro API and K3.1 Pro Highspeed with Built-in Tool Ecosystem (breakingnewsofficial) · platform.kimi.ai — Kimi.ai FunPlus · Tencent XtalPi K3.1 Pro · K3.1 Pro Highspeed · Kimi K3
- 2026-10-25: The State of Context Engineering: Failure Modes of Long Context (breakingnewsofficial) · dbreunig.com — O’Reilly AGNTCon Gemini 1.5 Gemini 1.5 Pro · Llama 4 Scout
- 2026-10-26: Benchmarks Are Surprisingly Compaction-Friendly: Evaluating Context Compaction in Coding Agents (breakingnewsofficial) · arxiv.org — GPT-5.3-CodeX
- 2026-10-26: EdgeRock-4B: 99% KV Cache Reduction with Fixed Memory for Long-Context Inference (breakingnewsofficial) · qbitai.com — Rock Matrix Hong Kong University of Science and Technology CMU Google EdgeRock-4B · Qwen3.5 4B · Gemma 4 E4B MiniCPM5 2B · LFM2.5-2.6B RWKV-7 G1j 13.3B AHN-GDN Qwen2.5 14B
- 2026-10-26: From Instruction-Following to Self-Learning: AI Agents, Recursive Self-Improvement, and the Human Oversight Imperative (breakingnewsofficial) · qbitai.com — Anthropic · OpenAI · MiniMax · Hugging Face Shanghai AI Lab Terence Tao · Claude Fable 5.1 · Claude Haiku 4.5 · GPT-6 Astra · GPT-6.1 Astra · GPT-5.6 Sol M3.1-Flash-Preview · Step 5 Preview AlphaGo
- 2026-10-26: Cross-Request KV Cache Exposure in GLM-5.3 on Cloudflare Workers AI (breakingnewsofficial) · blog.cloudflare.com — Cloudflare Knostic · z.ai · HackerOne Kevin Flansburg Michelle Chen Gadi Evron · GLM-5.3 · GLM-5.2
FAQ
What is Long Context?
Long Context refers to model and system designs that process large input windows while preserving useful attention and answer quality. GROUNDING tracks context limits, compression, memory, and retrieval-vs-context tradeoffs.
Which concepts are related to Long Context?
Related concepts tracked by the radar include Context Engineering, RAG.
What does this Long Context page track?
Dated updates, papers, and mentions of Long Context collected by the GROUNDING radar, most recently on 2026-10-26.