Thinking Machines Lab has released Inkling 2 Mini, a 26.4 billion parameter multimodal Looped Mixture-of-Experts model designed for local deployment on consumer hardware. The model supports a 1 million token context window and offers various quantization formats, including GGUF for CPU and Apple Silicon inference. Deployment notes highlight specific optimizations for NVIDIA Blackwell B200 GPUs and compatibility with vLLM, alongside API pricing for hosted inference. AI
IMPACT Enables local deployment of advanced multimodal models on consumer hardware, potentially lowering inference costs and increasing accessibility.
RANK_REASON Model release from a lab with a specific model name and version. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Apache Software License 2.0
- DeepSeek V4.1 Flash
- GGUF
- Inkling 2 Mini
- Nvidia B200
- Nvidia Blackwell B200
- Thinking Machines Lab
- TRT-LLM-gen
- Unsloth
- vLLM
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →