Models & Labs
Together AI launches Inkling model on day 0
Together AI Bloghigh confidence
Why it matters
- →Inkling's architecture supports efficient reasoning across multiple modalities.
- →Developers can leverage Inkling's capabilities without managing complex infrastructure.
- →The model's versatility allows for a wide range of applications, from scientific reasoning to audio comprehension.
Together AI has collaborated with Thinking Machines Lab to release the Inkling model on their platform from day one. Inkling is a multimodal mixture-of-experts model that handles text, image, and audio inputs, designed for efficient reasoning and broad task versatility. It features architectural innovations such as query-conditioned relative attention and short causal convolutions. This partnership allows developers to access Inkling's capabilities through Together AI's serverless platform, facilitating the development of applications with complex reasoning and multimodal input processing.
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.
