IBM launches Red Hat AI Inference and OpenShift Virtualization Service

IBM launched Red Hat AI Inference on IBM Cloud for real-time AI inferencing with models like Granite 4.0 and Llama 3.3, and Red Hat OpenShift Virtualization Service, both GA on May 22, 2026.

AMD acquires Taalas, a startup that bakes AI models directly into silicon

AMD is buying Canadian startup Taalas, which hard-codes model weights directly into inference chips. That makes them extremely fast but locks each chip to a single model. A demo chip hit over 16,000 tokens per second…

NVIDIA Introduces X-Token: Projection-Guided Cross-Tokenizer KD

NVIDIA researchers introduced X-Token, a projection-guided cross-tokenizer knowledge distillation method that improves performance on smaller models like Llama-3.2-1B.

Groq Launches Meta's Llama 3 Instruct AI Models on LPU™ Inference Engine

56% of sovereign AI LLMs are adapted from Llama, Qwen and Mistral

56% of Sovereign AI LLMs Are Adapted Models, With Meta’s Llama the Dominant Base, Followed by Alibaba’s Qwen and France’s Mistral Counterpoint Research

Limits max outputs of llama_context to save VRAM

llama: limit max outputs of `llama_context` (#23861) llama: save more VRAM by reserving n_outputs == n_seqs when possible add n_outputs_per_seq move n_outputs_max to server-context change ubatch to batch everywhere…