AI Models & LLM News: Releases, Benchmarks, Comparisons
New models, benchmarks and our tests
Inkling 2 Small: Thinking Machines' 594B MoE model scores 31 on AA
GPT-6 Cyber: OpenAI's Critical-risk model found 81 Firefox bugs
AI agents as teammates: 18 models tested, the best scores 82.2
Baidu ERNIE 5.5 Flash Base: 280B open weights lead 24 of 30 tests
Open-source AI grew tokens 22.5x but lost spend share
Kimi K3.1 Pro open weights: #1 open model in Agent Arena
Pangram Image 1.0 claims 99.2% detection at 0.27% false positives
A red heart image made Claude and GPT-5.4 write recursion every time
Tencent Hy4 leaves preview with vision and computer use
Gemini psychiatric intake: Google trains AI on simulated patients