Records found: 8
'Black Forest Labs released FLUX.2, a 32B flow-matching transformer for 4MP image generation and editing with multi-reference support, open dev weights, and production-ready integrations.'
'A concise technical comparison of vLLM, TensorRT-LLM, Hugging Face TGI v3 and LMDeploy, highlighting when to use each stack for production LLM inference based on throughput, latency and KV behavior.'
'Microsoft Research presents BitNet Distillation: a three‑stage pipeline that converts FP16 models to 1.58‑bit students, delivering up to 10× memory reduction and ~2.65× CPU speedup while preserving accuracy.'
'QeRL uses NVFP4 weight quantization plus LoRA and AQN to boost rollout throughput and exploration, allowing a 32B policy to be trained on a single H100 with competitive accuracy.'
'A hands-on tutorial showing how to export, optimize and quantize DistilBERT with Hugging Face Optimum and ONNX Runtime to speed up inference while keeping accuracy.'
'Explore how CUDA, ROCm, Triton and TensorRT map tensor programs to GPU hardware and which compiler-level optimizations yield the biggest performance gains.'
'Liquid AI unveils LFM2-VL, two open-weight vision-language models optimized for fast, low-latency on-device inference, offering 450M and 1.6B variants and easy integration via Hugging Face.'
'A technical deep dive into AI inference in 2025, detailing latency bottlenecks, optimization methods like quantization and pruning, and a roundup of the top nine inference providers.'