What are the hardware requirements for MAmmoTH Coder 34B?

MAmmoTH Coder 34B requires approximately 67.5 GB of VRAM and supports a context window of 16K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed MAmmoTH Coder 34B and how large is it?

MAmmoTH Coder 34B is developed by TIGER-Lab, a model with 34b 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 MAmmoTH Coder 34B?

MAmmoTH Coder 34B is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

MAmmoTH Coder 34B by TIGER-Lab — VRAM 67.5GB, 16K context

Name: MAmmoTH Coder 34B
Author: TIGER-Lab

MAmmoTH Coder 34B is an open-source language model by TIGER-Lab. Features: 34b LLM, VRAM: 67.5GB, Context: 16K, License: mit, Code Generating, LLM Explorer Score: 0.08.

Arxiv:2309.05653 Codegen Dataset:tiger-lab/mathinstruct En Endpoints compatible Llama Pytorch Region:us Sharded

Model Card on HF 🤗: https://huggingface.co/TIGER-Lab/MAmmoTH-Coder-34B

MAmmoTH Coder 34B Parameters and Internals

Model Type

Text generation, Math problem-solving

Use Cases

Applications:

Research, Educational software, Tutoring systems

Primary Use Cases:

General math problem-solving

Limitations:

Performance may vary based on complexity and specifics of the math problem

Considerations:

Not all mathematical fields may be comprehensively covered

Supported Languages

en (High)

Training Details

Data Sources:

MathInstruct Dataset

Methodology:

Hybrid Instruction Tuning with Chain-of-Thought and Program-of-Thought rationales

Input Output

Input Format:

Instruction-based with CoT or PoT rationale

Accepted Modalities:

text

Output Format:

Text - Solution to math problem

LLM Name	MAmmoTH Coder 34B
Repository 🤗	https://huggingface.co/TIGER-Lab/MAmmoTH-Coder-34B
Model Size	34b
Required VRAM	67.5 GB
Updated	2026-08-03
Maintainer	TIGER-Lab
Model Type	llama
Model Files	9.8 GB: 1-of-7 9.7 GB: 2-of-7 9.7 GB: 3-of-7 9.7 GB: 4-of-7 9.7 GB: 5-of-7 9.7 GB: 6-of-7 9.2 GB: 7-of-7
Supported Languages	en
Generates Code	Yes
Model Architecture	LlamaForCausalLM
License	mit
Context Length	16384
Model Max Length	16384
Transformers Version	4.31.0
Tokenizer Class	LlamaTokenizer
Beginning of Sentence Token	<s>
End of Sentence Token	</s>
Unk Token	<unk>
Vocabulary Size	32001
Torch Data Type	bfloat16

Rank the MAmmoTH Coder 34B 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

Quantized Models of the MAmmoTH Coder 34B

Model	Likes	Downloads	VRAM
MAmmoTH Coder 34B GGUF	2	618	14 GB
MAmmoTH Coder 34B GPTQ	2	10	17 GB
MAmmoTH Coder 34B AWQ	1	4	18 GB

Best Alternatives to MAmmoTH Coder 34B

Best Alternatives	Context / RAM	Downloads	Likes
...gpt 32K Codellama 34B Instruct	32K / 67.5 GB	56	2
CodeLlama 34B Instruct Hf	16K / 67.5 GB	22872	304
CodeLlama 34B Instruct Hf	16K / 1.4 GB	5	3
CodeLlama 34B Instruct Hf	16K / 1.4 GB	7	3
ReflectionCoder CL 34B	16K / 67.6 GB	7735	0
Speechless Codellama 34B V2.0	16K / 67.5 GB	96	17
CodeLlama 34B Hf	16K / 67.5 GB	27673	178
Phind CodeLlama 34B V2	16K / 67.5 GB	618	834
Phind CodeLlama 34B V1	16K / 67.5 GB	165	320
Phind CodeLlama 34B Python V1	16K / 67.5 GB	180	252

Note: green Score (e.g. "73.2") means that the model is better than TIGER-Lab/MAmmoTH-Coder-34B.