AI Models & LLM News: Releases, Benchmarks, Comparisons
New models, benchmarks and our tests
Meituan LongCat-2.5: a 1.6T agent model trained on Chinese chips
Gemini Notebook may add comic artifacts built from your sources
Phonic 1 tops the Speech to Speech Index by 0.2 points
OpenAI rewrites its Model Spec around five values, adds a root rule on oversight
Claude self-assessment flags 94% of urgent misuse in its own chats
Opus 5.5 built a better heist scene than Fable 5.1 for half the price
Mistral Large 4 goes open weight with 13 more points on Terminal-Bench
Index v5: open weights trail the leader by 26 points, not 12
Pika adds Kling 4.0 to its API at $0.157 per second
Kimi K3.1 Pro built a giant cat fight for $5.08, Sonnet cost 3x more