AMDROCmAI Software
ROCm 10 turns AI optimization into a developer workflow
AMD is pairing its updated compute stack with agent-assisted profiling, kernel optimization, diagnostics, and model-serving tools.
Source: AMD Newsroom
News // Signal over noise
AI News & Briefs
Concise updates on models, infrastructure, agents, and developer tooling—with original context and source attribution.
AI ComputeInferenceCustom Silicon
OpenAI publishes early results for its Jalapeño inference chip
OpenAI's first custom inference accelerator points to a future where model serving is optimized across silicon, software, and product workloads together.
Source: OpenAI
vLLMInferenceOpen Source
vLLM 0.26 expands model and quantization support
The v0.26.0 release adds an Inkling model support stack, Hopper attention work, speculative decoding support, LoRA, and NVFP4 quantization.
Source: vLLM GitHub Releases
AMD InstinctGPUHBM4
AMD's MI400 family targets both AI factories and scientific computing
The Instinct MI455X and MI430X split the next generation of AMD accelerators across frontier AI, sovereign infrastructure, and HPC workloads.
Source: AMD Newsroom
AnthropicAMD InstinctAI Infrastructure
Anthropic plans a multi-year expansion on AMD Instinct compute
A strategic agreement targeting up to two gigawatts of MI450-series deployments illustrates how frontier labs are diversifying accelerator supply.
Source: AMD Newsroom
OpenAIModelsAPI
OpenAI releases the GPT-5.6 model family
GPT-5.6 Sol, Terra, and Luna entered general availability, with later updates reducing API pricing for the family.
Source: OpenAI
OllamaLocal LLMGGUF
Ollama 0.30 improves GGUF compatibility and GPU performance
Ollama 0.30 integrates newer llama.cpp work, enables Vulkan by default, and reports higher throughput on supported NVIDIA hardware.
Source: Ollama Blog
Google CloudTPUAI Compute
Google splits its eighth-generation TPU strategy between training and inference
TPU 8t and TPU 8i use workload-specific designs for large-scale model training and latency-sensitive agent inference.
Source: Google Blog
AWS TrainiumvLLMSpeculative Decoding
AWS combines Trainium2 and vLLM for decode-heavy inference
An AWS engineering guide shows how speculative decoding can shift LLM serving economics when token generation dominates latency and cost.
Source: AWS Machine Learning Blog