Open Source
-
Research
Uno turns diffusion into a lossless speed layer for autoregressive LLMs
Uno combines autoregressive LLMs with diffusion adapters and Ψ-Spec verification, releasing code and checkpoints for parallel, distribution-preserving decoding.
-
Research
PatchBench shows why a stopped crash is not a verified AI security fix
PatchBench finds that single-PoC checks can overstate AI vulnerability-patching success and argues for independent security and semantic validation.
-
Business
NVIDIA confirms $12.93B Hugging Face deal and promises multi-vendor openness
NVIDIA formally confirms its Hugging Face deal and promises an open, multi-cloud, multi-accelerator platform. The real test will be product neutrality after closing.
-
Open Source
Token Monitor turns fragmented AI coding usage into one observability layer
Token Monitor unifies AI coding usage, cost and quota signals across tools, but its heterogeneous telemetry should not be treated as a billing source of truth.
-
Business
Nvidia reportedly agrees to buy Hugging Face for $12.9 billion
Nvidia reportedly agreed to acquire Hugging Face for $12.9 billion, a deal that could reshape open-model distribution and AI platform neutrality.
-
Open Source
FreeToken roadmap targets AMD, Apple Silicon and multi-GPU expansion
FreeToken's 2026 roadmap targets AMD ROCm, Apple Silicon, DGX Spark, multi-GPU, multimodal models and speculative decoding beyond its current NVIDIA-only support.
-
New Models
Tencent opens Hy4 preview for million-token coding and agent workflows
Tencent releases Hy4 preview with Apache 2.0 weights, 1M-token context, vLLM/SGLang deployment and a coding-focused 770B MoE architecture.
-
New Models
Fermion releases Phonon-1, a 415 MB open speech model for local transcription
Fermion releases Phonon-1, an Apache 2.0 English ASR model for local Apple Silicon and NVIDIA transcription with a 415 MB download.
-
New Models
AMALIA opens a 9B European Portuguese model with a 32K context and local deployment path
AMALIA releases a 9B open European Portuguese model with 32K context, SFT+DPO training, public evaluation resources and local vLLM serving.
-
New Models
Pipecat releases PhoneLLM, a 3.5B-active model tuned for voice agents
Pipecat's PhoneLLM Alpha 1 is an open 30B MoE with 3.5B active parameters, built for low-latency voice agents and evaluated with the new PhoneBench.
