Every daily briefing we have published. Use the calendar on any briefing page to jump to a specific day.
Real tests reveal up to 2.6x throughput variance for DeepSeek V4.1 Flash across providers; Qwen releases real-time interpretation model; Grok voice transcription at $0.10/hr; cache miss warnings.
Qwen released its first omni-modal model Qwen3.8-Omni-Flash with native audio-video understanding; Anthropic used Claude to optimize and open-source 30+ biology models with an average 4x speedup.
OpenAI offers free GPT-6 Astra Pro to verified US clinicians; Google DeepMind opens AlphaGenome Atlas for disease genetics. StepFun also releases stepaudio3 music generation model.
Today in AI, StepFun launched the StepAudio 3 audio model family with API access, showcasing strong real-time speech capabilities; Perplexity open-sourced its CobbleDB database, slashing read latency and costs. Several free models and tools also debuted.
SiliconFlow launches Hy4, a 770B-parameter open-source model with pricing; MiniMax H3 generates 14.4s of 768p video in 9.0s on 8x B200, exceeding 2x real-time. Also: Qoder's 500 daily free credits reset, Hugging Face AI-agent attack, and more.
Command Code launches DeepSeek V4.1 Flash, adding $60 usage to GOAT plan (valid 10 days) for 6x weekly usage; MiniMax H3 video model gets community optimization—FastH3 achieves 4-step distillation and runs on Apple Silicon.
OpenAI endorses Anthropic's call to pause frontier AI development and commits to giving independent evaluators system access. Meanwhile, disk-backed caching extends AI cache token lifetime beyond 24 hours, potentially cutting inference costs.
