A custom fine-tuned AI model developed by Bridgewater AIA Labs and Thinking Machines Lab outperforms GPT, Claude, and Gemini on financial tasks, achieving 84.7% accuracy versus frontier models' ~50%, cutting errors by 30% and slashing inference costs by 13.8x.
Anthropic's Claude Cookbooks GitHub repository has surpassed 46,000 stars, offering developers a growing collection of copy-paste Python code snippets covering AI capabilities like text classification, multimodal vision, and retrieval augmented generation, with 81 contributors and active development as of June 2026.
AI inference chip startup Etched surges to a $5 billion valuation after closing a $500 million funding round, backed by Peter Thiel, Geoffrey Hinton, and Fei-Fei Li, while securing $1 billion in contract orders as it takes direct aim at Nvidia's dominance in the booming AI chip market.
Wix-owned vibe-coding platform Base44 launches Base1, a custom AI model trained on tens of millions of real user interactions, aiming to outperform competitors in speed and efficiency as the vibe-coding market heats up with rivals like Lovable hitting $500 million in ARR and frontier labs like Anthropic encroaching on the …
Moondream's new Photon inference engine eliminates GPU idle time through pipelined decoding, boosting vision-language model throughput by up to 35% on NVIDIA B200 hardware and achieving near-realtime inference at approximately 33ms.
Google launches Nano Banana 2 Lite and Gemini Omni Flash, delivering lightning-fast, low-cost AI image and video generation for developers and consumers, with images produced in just 4 seconds and video output available at $0.10 per second.
Anthropic launches an internal drug discovery program targeting neglected diseases and unveils Claude Science, a new AI tool for drugmakers, positioning itself as a hands-on healthcare player alongside tech giants like Alphabet, Apple, and Amazon.
Google's Gemini Omni Flash leads a fiercely competitive Text-to-Video AI leaderboard with a top score of 1527, but Chinese tech giants ByteDance and Alibaba are closing the gap fast, claiming second and third place respectively among 41 competing models.
DeepSeek releases DSpark, a free open-source AI framework that boosts large language model inference speeds by up to 85% using confidence-scheduled speculative decoding, delivering over 50% throughput gains in live production and supporting popular open-weight models like Qwen and Gemma.
DeepSeek AI releases DeepSpec, an open-source speculative decoding framework supporting algorithms like Eagle3 and DSpark, complete with pre-trained checkpoints for models including Qwen3 and Gemma-4-12B now available on Hugging Face under the MIT License.