license: apache-2.0
datasets:
- continuedev/instinct-data
Instinct, the State-of-the-Art Open Next-Edit Model
This repo contains the model weights for Instinct, Continue's state-of-the-art open next-edit model. Robustly fine-tuned from Qwen2.5-Coder-7B on our dataset of real-world code edits, Instinct intelligently predicts your next move to keep you in flow.
Serving the model
Ollama: We've released a Q4_K_M GGUF quantization of Instinct for efficient local inference. Try it with Continue's Ollama integration.
Besides Ollama, there are many ways to plug a local model into Continue; we internally used an endpoint served by SGLang, which is one of the options below. Quantizing for faster inference is also an option that worked well for us. Serve the model using either of the below options, then connect it with Continue.
SGLang: python3 -m sglang.launch_server --model-path continuedev/instinct --load-format safetensors
vLLM : vllm serve continuedev/instinct --served-model-name instinct --load-format safetensors
Learn more
For more information on the work behind Instinct, please refer to our blog.