What are the hardware requirements for Macaron V1 Tall?

Macaron V1 Tall requires approximately 72.2 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Macaron V1 Tall and how large is it?

Macaron V1 Tall is developed by mindlab-research, a model with 35b 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 Macaron V1 Tall?

Macaron V1 Tall is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Macaron V1 Tall by mindlab-research — VRAM 72.2GB

Name: Macaron V1 Tall
Author: mindlab-research

Macaron V1 Tall is an open-source language model by mindlab-research. Features: 35b LLM, VRAM: 72.2GB, License: mit, LLM Explorer Score: 0.31.

A2ui Base model:finetune:qwen/qwen3... Base model:qwen/qwen3.6-35b-a3... Coding-agent Conversational En Endpoints compatible Eval-results Generative-ui Image-text-to-text Macaron Macaron-v1 Mixture-of-lora Personal-agent Qwen3.6 Qwen3.6-35b-a3b Qwen3 5 moe Region:us Safetensors Sharded Tensorflow Tool-use Ui4a Zh

Model Card on HF 🤗: https://huggingface.co/mindlab-research/Macaron-V1-Tall

Macaron V1 Tall Parameters and Internals

LLM Name	Macaron V1 Tall
Repository 🤗	https://huggingface.co/mindlab-research/Macaron-V1-Tall
Base Model(s)	Qwen/Qwen3.6-35B-A3B Qwen/Qwen3.6-35B-A3B
Model Size	35b
Required VRAM	72.2 GB
Updated	2026-08-09
Maintainer	mindlab-research
Model Type	qwen3_5_moe
Model Files	4.0 GB: 1-of-26 1.3 GB: 2-of-26 3.4 GB: 3-of-26 3.4 GB: 4-of-26 3.4 GB: 5-of-26 4.0 GB: 6-of-26 1.1 GB: 7-of-26 4.0 GB: 8-of-26 1.1 GB: 9-of-26 4.0 GB: 10-of-26 1.1 GB: 11-of-26 3.4 GB: 12-of-26 1.6 GB: 13-of-26 3.4 GB: 14-of-26 1.6 GB: 15-of-26 4.0 GB: 16-of-26 1.1 GB: 17-of-26 4.0 GB: 18-of-26 1.1 GB: 19-of-26 3.4 GB: 20-of-26 1.6 GB: 21-of-26 3.4 GB: 22-of-26 3.4 GB: 23-of-26 3.4 GB: 24-of-26 3.8 GB: 25-of-26 2.2 GB: 26-of-26
Supported Languages	en zh
Model Architecture	Qwen3_5MoeForConditionalGeneration
License	mit
Model Max Length	262144
Transformers Version	4.57.1
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Errors	replace

Rank the Macaron V1 Tall 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 Macaron V1 Tall

Best Alternatives	Context / RAM	Downloads	Likes
....5 35B A3B Sceptic V3 1M W4A16	1024K / 26.1 GB	223	0
...AgentWorld 35B A3B Abliterated	256K / 69.2 GB	962	18
Ornith 1.5 35B A3B NPU2	256K / GB	96	1
Qwen3.6 35B A3B NPU2	256K / GB	107	0
Qwen3.6 35B A3B	0K / 72.2 GB	5540855	2648
Qwen3.5 35B A3B	0K / 72.3 GB	2158511	1482
Qwen3.6 35B A3B NVFP4	0K / 23.4 GB	8788684	482
Ornith 1.0 35B	0K / 70.4 GB	1465715	506
Qwen3.6 35B A3B FP8	0K / 0.8 GB	8984811	343
Qwen AgentWorld 35B A3B	0K / 69.2 GB	66155	665

Note: green Score (e.g. "73.2") means that the model is better than mindlab-research/Macaron-V1-Tall.