What are the hardware requirements for Ziya Coding 15B V1?

Ziya Coding 15B V1 requires approximately 31.1 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Ziya Coding 15B V1 and how large is it?

Ziya Coding 15B V1 is developed by IDEA-CCNL, a model with 15b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Ziya Coding 15B V1?

Ziya Coding 15B V1 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Ziya Coding 15B V1 by IDEA-CCNL — VRAM 31.1GB

Name: Ziya Coding 15B V1
Author: IDEA-CCNL

Ziya Coding 15B V1 is an open-source language model by IDEA-CCNL. Features: 15b LLM, VRAM: 31.1GB, License: gpl-3.0, Code Generating, LLM Explorer Score: 0.08.

Arxiv:2210.08590 Codegen En Endpoints compatible Gpt bigcode Pytorch Region:us Sharded Zh

Model Card on HF 🤗: https://huggingface.co/IDEA-CCNL/Ziya-Coding-15B-v1

Ziya Coding 15B V1 Parameters and Internals

Model Type

text-generation

Use Cases

Areas:

Research, Development, Software, AI

Applications:

Code generation, Code modification, Code interpretation, NL2SQL

Primary Use Cases:

Generating and modifying code, Code interpretation, Code continuation, Instruction following

Additional Notes

Cites detailed performance metrics achieved on HumanEval and MBPP benchmarks, showing strong results post-fine-tuning.

Supported Languages

zh (high), en (high)

Training Details

Data Sources:

100B tokens of high-quality corpus including Chinese, English, and code

Data Volume:

100B tokens

Methodology:

Continual pretraining, Supervised finetuning

Training Time:

10 days

Hardware Used:

144 A100 GPUs with 40GB each

Model Architecture:

StarCoderBase

Input Output

Input Format:

Text prompt

Accepted Modalities:

text

Output Format:

Generated code and text

LLM Name	Ziya Coding 15B V1
Repository 🤗	https://huggingface.co/IDEA-CCNL/Ziya-Coding-15B-v1
Model Size	15b
Required VRAM	31.1 GB
Updated	2026-08-02
Maintainer	IDEA-CCNL
Model Type	gpt_bigcode
Model Files	3.0 GB: 1-of-11 2.7 GB: 2-of-11 2.7 GB: 3-of-11 3.0 GB: 4-of-11 3.0 GB: 5-of-11 2.7 GB: 6-of-11 2.7 GB: 7-of-11 3.0 GB: 8-of-11 3.0 GB: 9-of-11 2.7 GB: 10-of-11 2.6 GB: 11-of-11
Supported Languages	zh en
Generates Code	Yes
Model Architecture	GPTBigCodeForCausalLM
License	gpl-3.0
Model Max Length	2048
Transformers Version	4.32.0
Tokenizer Class	GPT2Tokenizer
Beginning of Sentence Token	<\|endoftext\|>
End of Sentence Token	<\|endoftext\|>
Unk Token	<\|endoftext\|>
Vocabulary Size	49155
Torch Data Type	bfloat16
Activation Function	gelu
Errors	replace

Rank the Ziya Coding 15B V1 capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Ziya Coding 15B V1

Best Alternatives	Context / RAM	Downloads	Likes
WizardCoder Guanaco 15B V1.1	0K / 31 GB	52	12
WizardCoder Guanaco 15B V1.0	0K / 31 GB	55	5
MoTCoder 15B V1.0	0K / 30.9 GB	26	2
WizardCoder 15B V1.0	0K / 31 GB	211	761
Interplay AppCoder	0K / 31.2 GB	10	2
Codes 15B Spider	0K / 62.1 GB	19	0
Codes 15B Bird With Evidence	0K / 62.1 GB	5	0
WizardCoderSQL 15B V1.0	0K / 31.2 GB	6	1
Codes 15B	0K / 62.1 GB	11	9
...zardCoder 15B V1.0 Sharded 8gb	0K / 62.2 GB	18	0

Note: green Score (e.g. "73.2") means that the model is better than IDEA-CCNL/Ziya-Coding-15B-v1.