The repository is ungated and provides model weights in safetensors format.
The repository metadata lists 39,785,287,456 BF16 parameters and a total weight-file size of 79,578,467,461 bytes. It was created on October 21, 2026 and modified the same day. No hosted inference provider is listed on the page.
The supplied instructions cover loading Flint with Transformers, serving it through vLLM or SGLang with OpenAI-compatible chat-completions endpoints, and running it with Docker Model Runner. The page also points to compatible quantizations for llama.cpp, Ollama, LM Studio, or similar applications.
The source does not establish benchmark results, coding-task performance, hardware requirements, training data, or a 2.5B active-parameter figure. It also does not state a price or a managed API access plan.