CherryPit-1.1-2B-WebDev-Preview-1024-GGUF

⚠️ Preview release. Results and weights may change before the final CherryPit-1.1.

CherryPit-1.1-2B-WebDev-Preview is the next version of CherryPit from DRUPEs (Dense Representation & Unified Parameter Encoding Systems), built on MiniCPM5-2B. Its focus is building web apps: given a set of tests, it writes React components that pass them far more often than CherryPit-1.0, while keeping general coding and math level with or better than 1.0.

Results vs CherryPit-1.0

Both models were evaluated with the same setup.

Model Details

• Developed by: DRUPEs
• Model type: Causal language model
• Parameters: ~2B
• Base model: MiniCPM5-2B
• Previous version: CherryPit-1.0-2B-GGUF
• Language(s): English
• License: Apache 2.0
• Version: 1.1 WebDev Preview
• Checkpoint: 1024 (October 24, 2026)

How to Use

Ollama

ollama run hf.co/DRUPEs/CherryPit-1.1-2B-WebDev-Preview-1024-GGUF:Q4_K_M

LM Studio

Search for DRUPEs/CherryPit-1.1-2B-WebDev-Preview-1024-GGUF in the Discover tab.

llama.cpp

llama-cli -hf DRUPEs/CherryPit-1.1-2B-WebDev-Preview-1024-GGUF:Q4_K_M

Files

More quantizations will come with the final 1.1 release.

Limitations

• This is a preview. Expect changes before the final 1.1 release.
• The biggest gains are in React and web app generation. General coding is roughly level with 1.0: MBPP+ is up, while LiveCodeBench and HumanEval+ are about the same or slightly lower.
• CherryPit is a 2B model: it can make factual errors and write buggy code, so review its output before relying on it.
• Evaluations were run in English.

Citation

@misc{drupes2026cherrypit11,
  title  = {CherryPit-1.1-2B-WebDev-Preview},
  author = {DRUPEs},
  year   = {2026},
  url    = {https://huggingface.co/DRUPEs/CherryPit-1.1-2B-WebDev-Preview-1024-GGUF}
}