Models & Labs
Together AI Introduces Autoscaling for LLM Inference
Together AI Bloghigh confidence
Why it matters
- →Autoscaling based on inference-native metrics optimizes resource use for LLM deployments.
- →It addresses the limitations of traditional autoscaling methods in handling LLM traffic.
- →This feature is crucial for managing costs and performance in GPU-constrained environments.
Together AI has launched a new autoscaling feature for large language model (LLM) inference on its platform. This feature allows deployments to scale based on specific metrics such as in-flight requests and GPU utilization, addressing the unique challenges of LLM serving. Traditional autoscaling methods often fail due to misleading metrics and lengthy cold starts, but Together AI's approach offers a catalog of inference-native metrics for more precise scaling. This advancement helps optimize resource allocation, especially in environments where GPU resources are limited.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
More in Models & Labs
Models & Labsother
llama.cpp b12340 adds CUDA 13 and ROCm 10.0
This release quietly closes the hardware gap for local inference by adding default support for CUDA 13 and ROCm 10.0 across Linux and Windows. Users with newer NVIDIA GPUs or AMD accelerators no longer need to compile from source to get performance, while Apple Silicon builds now ship KleidiAI disabled by default, shifting the focus back to standard ARM optimizations. The inclusion of Snapdragon support for Linux and Android marks a significant step toward on-device AI beyond the x86/ARM desktop ecosystem. It is less about new features and more about ensuring llama.cpp remains the universal runtime for every available accelerator.
llama.cpp Releases
Models & Labscoding
llama.cpp b12342 adds few-row MMQ for speculative decoding
Speculative decoding just got a significant speed boost on NVIDIA GPUs. This release introduces a specialized kernel for 'few-row' matrix multiplications, targeting the small batch sizes typical of verification passes in speculative decoding workflows. By optimizing these specific operations for quantized formats like Q8_0 and IQ4 variants, llama.cpp reduces latency where it matters most during token generation. The change is CUDA-only for now but leverages hardware capabilities from Ampere onwards, making local inference faster without requiring new model architectures or complex configuration changes.
