JetBrains Open Sources Mellum2: 12B Parameter AI Coding Model
JetBrains releases Mellum2, a 12B parameter MoE model for code completion, offering high performance with low computatio…
📂 LLM News
Latest articles in LLM News
Alibaba Unveils Qwen 3.7 Plus: A Cost-Effective AI Challenger
Alibaba releases Qwen 3.7 Plus, a mid-tier model designed to balance performance and cost for global developers.
Microsoft Unveils MAI-Thinking-1 and Copilot Super App
Microsoft reveals its first self-developed reasoning model MAI-Thinking-1 and a new Copilot super app at Build 2026, sig…
Qwen3.7-Plus: New Multimodal Agent AI
Alibaba's Qwen3.7-Plus launches as a powerful multimodal agent, enhancing complex reasoning and visual analysis for ente…
Speed Up LLM Loading with AWS GPUDirect
New AWS integrations cut LLM load times by 50% using GPUDirect and FSx for Lustre.
MuskAI Unveils Ultra-Low Cost GPT-5.5 API Access
New MuskAPI service offers GPT-5.5 and Codex models at unprecedented low rates with high stability for developers.
Chinese AI Coding Models: Ready for Production?
Domestic Chinese LLMs now rival Western tools in coding productivity, offering viable alternatives to GitHub Copilot for…
MiniMax M3: Open-Weight AI Challenges Giants
Chinese firm MiniMax launches M3, an open-weight model with 1M token context and native multimodality.
Claude's Glitch: Bypassing Hourly Limits
Users discover a loophole in Anthropic's Claude AI, allowing unlimited usage despite strict hourly caps.
Claude 5-Hour Limit Bug: Unlimited Output?
Users report Claude bypassing 5-hour rate limits, allowing continuous output without reducing weekly quotas.
MiniMax M3: Long Context & Agent Power
MiniMax launches M3, targeting enterprise agents with 2M token context and tool execution.
Kog AI Engine Hits 3000 Tokens/s on Standard GPUs
Kog AI releases KIE, achieving 3000 tokens/s on AMD MI300X without quantization or speculative decoding.