CodeGemma
GoogleGoogle's code-specialized Gemma family: a 2B pretrained model for fast fill-in-the-middle completion, a 7B pretrained model for infilling and natural language, and a 7B instruction-tuned variant for code chat. All are further-trained on 500 billion to a trillion tokens of public code, mathematics and synthetic data.
The training data and recipe were not released, and the repositories require an access request, so openness was read from the model card rather than a download. Largely superseded for coding by later families. Verified 2026-08-13 via the model card and Hugging Face's CodeGemma release post.
Openness
3 high confidence- weights
- open
- data
- closed
- code
- closed
- license
- Gemma-Terms-of-Use(custom,use-restricted,non-OSI)
Open weights under Google's custom Gemma terms, with no open training data or code. A license that bounds commercial use caps at 3, where one forbidding commerce outright caps at 2; Gemma's terms are the former, so this lands at 3. No sibling record offers a comparison on the same license, since the surviving Gemma record carries an Apache-2.0 license rather than the Gemma terms.
- https://huggingface.co/google/codegemma-7b-it recorded 2026-08-13
`license:gemma` in the repo tags and `"gated":"manual"`, an access request under the Gemma terms; sizes and intended use documented; no training corpus and no training code published
Adoption
2 medium confidenceThe five CodeGemma repositories the tier covers sum to 28,295 Hugging Face downloads in the trailing 30 days (google/codegemma-7b-it 4,573; google/codegemma-7b 4,262; google/codegemma-2b 19,078; google/codegemma-1.1-7b-it 50; google/codegemma-1.1-2b 332), which bands at level 2 (10K-100K). All five are declared on this product, and adoption is summed across the releases a tier covers.
- https://huggingface.co/api/models/google/codegemma-7b-it recorded 2026-08-12
4,573 downloads in the trailing 30 days for google/codegemma-7b-it
- https://huggingface.co/api/models/google/codegemma-7b recorded 2026-08-12
4,262 downloads in the trailing 30 days for google/codegemma-7b
- https://huggingface.co/api/models/google/codegemma-2b recorded 2026-08-12
19,078 downloads in the trailing 30 days for google/codegemma-2b
- https://huggingface.co/api/models/google/codegemma-1.1-7b-it recorded 2026-08-12
50 downloads in the trailing 30 days for google/codegemma-1.1-7b-it
- https://huggingface.co/api/models/google/codegemma-1.1-2b recorded 2026-08-12
332 downloads in the trailing 30 days for google/codegemma-1.1-2b
Capability
3 medium confidenceA 2024 code-completion family, competent for its size and generation and well below the 2026 coders. The Hugging Face release post's comparison table puts CodeGemma 7B at 40.13 on HumanEval Python against DeepSeek Coder 7B's 45.83 and StarCoder 2 15B's 44.15, on 500B tokens of extra training, which is the competent-not-leading position the 3 records.
- https://github.com/huggingface/blog/blob/main/codegemma.md recorded 2026-08-13
"CodeGemma-7B outperforms similarly-sized 7B models except DeepSeek-Coder-7B on HumanEval"; comparison table - CodeGemma 7B Python 40.13 / JavaScript 43.06, DeepSeek Coder 7B 45.83 / 45.9, StarCoder 2 15B 44.15 / 44.24, CodeGemma 2B 27.28 / 29.94; 2B base for infilling, 7B base, 7B instruct
Verified 2026-08-12