https://myprivateclaw.com/edge/deepseek-v4-1-flash-kv-cache-compression MyPrivateClaw en 2026-09-18T12:00:00.000Z DeepSeek AI publishes DeepSeek-V4.1-Flash, a multimodal MoE model built around KV cache compression https://myprivateclaw.com/edge/flyweight-v0-2-0-first-release MyPrivateClaw en 2026-09-18T12:00:00.000Z Flyweight publishes its first public release, v0.2.0, an open-source native C++/CUDA GGUF inference runtime for MoE models https://myprivateclaw.com/edge/aspire-async-batched-self-speculative-decoding MyPrivateClaw en 2026-09-17T12:00:00.000Z ASPIRE, submitted on 16 Sep 2026, proposes non-synchronized batched self-speculative decoding for long-context LLM inference https://myprivateclaw.com/edge/ollama-v0-34-2-release MyPrivateClaw en 2026-09-17T12:00:00.000Z Ollama v0.34.2 fixes MLX speculative decoding memory growth and adds first-run setup