#Large Language Models
AIAlibaba Unveils Qwen3.8-Max, Its Most Powerful Open AI Model Yet
Alibaba has announced Qwen3.8-Max, its largest AI model to date. The open-weight model supports text, images, and video while introducing a 1 million-token context window and improved efficiency.
AIDeepSeek V4-Flash API Hits Public Beta With Agent Benchmarks That Beat V4-Pro-Preview
DeepSeek V4-Flash API entered public beta on July 31 with agent benchmarks that beat V4-Pro-Preview on every task and native Codex compatibility.
AIClaude Opus 5 vs Fable 5: Six Benchmarks, Side-by-Side Pricing, and When to Switch
Opus 5 lands within 0.5% of Fable 5 on CursorBench at max effort, beats it on Frontier-Bench, and costs half as much at $5/$25 per million tokens.
AIClaude Opus 5 Launches: Same Price as Opus 4.8, Within 0.5% of Fable 5 on Coding Benchmarks
Anthropic launched Claude Opus 5 at $5/$25 per million tokens - Opus 4.8's price - while landing within 0.5% of Fable 5 on CursorBench at half the cost.
AIGoogle Launches Three New Gemini Models and Teases Gemini 4 While Its Flagship Sits in Limbo
Google shipped three Flash-class Gemini models on Tuesday and dropped its first hint about Gemini 4. Gemini 3.5 Pro is still not out.
AIGoogle Is Building a Chip With Gemini Baked Directly Into the Hardware
Google is building a chip codenamed "Frozen v2" that bakes Gemini directly into hardware - 6-10x more efficient than its current AI chips, targeting 2028.
AIAnthropic Pays $1.5 Billion to Close AI Copyright Case - Fair Use Won, Piracy Lost
A federal judge approved the Anthropic copyright settlement for $1.5 billion on July 20 - the largest in US history. Training AI on books is fair use. Downloading pirated ones from Library Genesis was not.
SecurityHugging Face Hacked by an Autonomous AI Agent - US LLMs Blocked Defenders, So GLM 5.2 Ran the IR
Hugging Face discloses a July 2026 security incident: autonomous AI agent attack, US LLM guardrails blocked forensics, GLM 5.2 ran the IR on-prem.
AIKimi K3 Pauses New Subscriptions After 48 Hours of Demand Maxed GPU Capacity
Kimi K3 hit capacity limits after 48 hours. New subscriptions paused; existing users unaffected. Moonshot splits into Kimi Membership and Kimi Code Membership.
AIQwen3.8 Goes Live With 2.4T Parameters - Alibaba Claims Second to Fable 5, Open Weights Coming
Alibaba's Qwen3.8 preview lands with 2.4T parameters and multimodal support. Alibaba claims second to Fable 5. Open weights are promised but have no date.
AIKimi K3 vs Claude Fable 5: Cheaper and Faster on Agents, Behind on Professional Work
K3 wins SWE Marathon, BrowseComp, and Terminal Bench at 70% lower cost. Fable 5 leads intelligence and professional work. Pick based on your task.
AINVIDIA Nemotron 3 Embed Hits #1 on RTEB - Three Open Models for Production RAG and Agentic Retrieval
NVIDIA's Nemotron 3 Embed ranks #1 on RTEB. Three open commercial models - 8B hits 78.46 NDCG@10, 1B hits 72.38 - with 32K context and code retrieval training.