Moonshot AI

Latest AI news, models and releases from Moonshot AI. ['GLM 5.2', 'GLM-5.2', 'Hermes', 'K2', 'K2.6', 'K3', 'K3-mini', 'K3-MoE', 'Kimi', 'Kimi 2.6', 'Kimi-2.6', 'Kimi 3', 'Kimi 3T', 'Kimi Hosted Agent', 'Kimi k1.5', 'Kimi K1.5', 'kimi-k2', 'Kimi K2', 'kimi-k2-0711-preview', 'Kimi K2.5', 'Kimi-K2.5-1T-A32B', 'Kimi K2.6', 'Kimi K2.7', 'Kimi K2.7 Code', 'Kimi K2 Thinking', 'Kimi K3', 'Kimi-K3', 'Kimi K3.5', 'Kimi K4', 'Kimi K5', 'Kimi Linear', 'Kivine', 'Krea 2', 'Krea 2 Turbo', 'Mamba', 'Mamba-2', 'Mamba-3', 'MiniCPM-RobotManip', 'MiniCPM-RobotTrack', 'Minimax']

Official site ↗

Models 🇺🇸

Fable Writes GPU Cores, AI Automation and Analog Computing: Import AI 464 Digest

Fable created the fastest mega-kernel for KernelBench-Mega, achieving an 18.71x speedup. Researchers from CAIS and Scale Labs documented an increase in AI agent success rates on online freelancing from 2.5% (October 2025) to 16.1% (July 2026). The OSWORLD 2.0 benchmark for evaluating long-duration computer tasks (median 1.6 hours) has been released; the best result is 20.6%. JD published details of its Oxygen AIIC inventory management system, running on Huawei Ascend NPUs.

OpenAI

Anthropic

Moonshot AI

MiniMax

Alibaba/Qwen

JD.com

Import AI27.07 · 04:07

Research 🇷🇺

Finam AI Lab Updates Financial Benchmark FINESSE-Bench for LLMs

Finam's Artificial Intelligence Laboratory has released an updated version of the financial benchmark FINESSE-Bench. The new version adds a technical analysis dataset CFTe-like Level 1, fixes 169 CFA-like Level 1 questions, improves metric calculation with bootstrapping, and expands the model pool to 33 in comparative tables.

Anthropic

Moonshot AI

Google/DeepMind

MiniMax

OpenAI