Model

CodeGemma 7B

OPEN

Google · CodeGemma family · released Apr 9, 2024

Gemma 7B continued-pretrained on ~500B tokens of code (80/20 code-to-text) with fill-in-the-middle for completion.

ReasoningCodingVisionFunction callingTool useAgentic

252.0

Elo · rank #355

Parameters

8.54B

Active params

8.54B (dense)

Context

8K tokens

Architecture

Dense decoder-only transformer (Gemma 7B architecture, continued pretrain for code)

License

Gemma Terms of Use

Languages

—

API price (in/out)

No hosted API

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

GSM8KMath44.2%#141

best: Llama 3.1 405B · 96.8%

HumanEval FIMCoding76.1%#4

best: Codestral 22B · 91.6%

HumanEvalCoding44.5%#131

best: Claude Opus 4.5 · 99.4%

MATH-500Math19.9%#183

best: GPT-5 · 99.4%

MBPPCoding56.2%#63

best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%

Run it locally

VRAM @ Q4

6 GB

VRAM @ FP16

18 GB

Fits on (Q4)

RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Throughput data unavailable.

Quantizations

GGUF Q4_K_M · GGUF Q8_0 · AWQ · MLX

CodeGemma family

Elo progression across releases

CodeGemma 7BApr 2024252.0

Resources

HFgoogle/codegemma-7b GHgoogle-deepmind/gemma DOCSai.google.dev/gemma/docs/codegemma/model_card APIAvailable via Google API

API price weights · each benchmark row carries its own source badge (see methodology)