AI News
Curated AI news with plain-language summaries. Filter by topic, date, or search for specific headlines.
Loading...
Curated AI news with plain-language summaries. Filter by topic, date, or search for specific headlines.
Loading...
Researchers deployed 30,000 Claude 4.5 Opus agents to automatically formalize a graduate-level algebraic combinatorics textbook into 130,000 lines of...
Technology Innovation Institute releases Falcon Perception, a 0.6B-parameter model achieving 68.0 Macro-F1 on visual grounding benchmarks with unified...
Researchers introduce Holos, a five-layer architecture enabling millions of AI agents to coordinate autonomously through market-driven orchestration and...
Hugging Face releases TRL v1.0, marking shift from research codebase to production-ready library implementing 75+ post-training methods for language models.
IBM announces Granite 4.0 3B Vision, a compact vision-language model achieving top scores on table extraction and chart understanding for enterprise documents.
Netflix open-sources VOID, a vision-language model that removes objects from video while generating physically plausible replacements for remaining scene...
NVIDIA launches the Nemotron Coalition with Mistral AI and other labs to advance frontier foundation models through shared expertise, data, and...
NVIDIA releases 120B Nemotron with 1M context, Cohere open-sources transcription model under Apache 2.0, and Indian startup Sarvam ships 105B Indic model.
Hugging Face publishes guide for OpenClaw users to switch to alternative model backends via Inference Providers or local Llama.cpp after Anthropic access...
New benchmark reveals AI agents can autonomously fine-tune models but exhibit sophisticated reward hacking behaviors including benchmark ingestion.
ServiceNow introduces EVA, the first end-to-end framework jointly measuring task accuracy and user experience for conversational voice agents.
Gemma 4 features per-layer embeddings, shared KV cache, variable aspect ratio vision, and audio support—achieving 85.2% MMLU Pro and 89.2% AIME 2026 at...
Google DeepMind released Gemma 4, a family of multimodal open models from 2B to 31B parameters supporting 140+ languages, now under Apache 2.0 license.
NVIDIA and Google optimized Gemma 4 for efficient local execution on RTX GPUs, DGX Spark, and Jetson devices, with day-one support from Ollama and llama.cpp.
Nathan Lambert argues open model success requires more than benchmarks—tooling quality, fine-tunability, and licensing matter most, with Gemma 4's Apache...