Dots3-Note Preview
- Context
- 1Mtokens
- Released
- 3mo ago
- Size
- 753Bparameters
Macaron-V1-Venti is Mind Lab's flagship text-generation model in the Macaron-V1 family, following Macaron-V1-Preview and described as the first model post-trained on a GLM-5.2 base. Its launch highlights a Mixture-of-LoRA design that routes each request to one of four specialist adapters for chat, agent tasks, coding, or generative UI. Mind Lab positions the model for personal intelligence, tool use, coding, and code-native generative UI work, with post-training aligned to its agentic serving harness.
- Language model
- Open weights · 753B
- 1M context
- Reasoning
- Tools
How it compares
Additional source evidence
- Source model
- mindlab/macaron-v1-venti
- Last updated
- 2026-07-21
- Model context
- 1M
- Model output limit
- 131K
- Open weights
- Yes
- File attachments
- No
- Reasoning
- Yes
- Tool calling
- Yes
- Temperature control
- Yes
Downloadable weights
Complete source record (JSON)
{
"id": "mindlab/macaron-v1-venti",
"name": "Macaron V1 Venti",
"description": "GLM-5.2-based Mixture-of-LoRA model with specialists for chat, personal agents, coding, and generative UI",
"last_updated": "2026-07-21",
"attachment": false,
"reasoning": true,
"tool_call": true,
"temperature": true,
"modalities": {
"input": [
"text"
],
"output": [
"text"
]
},
"limit": {
"context": 1048576,
"output": 131072
},
"release_date": "2026-07-21",
"open_weights": true,
"license": "MIT",
"weights": [
{
"label": "Hugging Face",
"url": "https://huggingface.co/mindlab-research/Macaron-V1-Venti"
}
]
}Frequently asked questions
Running locally
Macaron V1 Venti needs an estimated 525.7 GB of VRAM at Q4, 921 GB at Q8 and 1,733.2 GB at FP16. These are estimates from the 753B parameter count, weights plus about 15% and 0.5 GB for a short context.
No. Macaron V1 Venti does not fit on a 24 GB, 16 GB, 12 GB or 8 GB GPU at Q4, Q8 or FP16 (estimates).
Privacy and data
Yes. You can self-host Macaron V1 Venti for full control of your data because it has open weights.
Capabilities and specs
The context window of Macaron V1 Venti is 1M tokens (1,048,576). Its maximum output is 131K tokens (131,072).
No. Macaron V1 Venti only accepts text input, so it cannot read images, audio, video or PDFs.
Yes. Macaron V1 Venti supports tool calling (function calling).
Yes. Macaron V1 Venti supports a reasoning mode.
Quality and benchmarks
The closest alternatives to Macaron V1 Venti, matched on model kind and release date with no comparable price, are Dots3-Note Preview by Dots Studio, Kolibri 1 by Aleph Alpha and Nemotron 3.5 Super 120B A12B by NVIDIA.
Dots3-Note PreviewMacaron V1 Venti vs Dots3-Note PreviewKolibri 1Macaron V1 Venti vs Kolibri 1Nemotron 3.5 Super 120B A12BMacaron V1 Venti vs Nemotron 3.5 Super 120B A12B
Licensing and access
Macaron V1 Venti has open weights, not full open source. Its license is MIT.
Yes. Macaron V1 Venti is licensed under MIT, a permissive license that allows commercial use.
The Macaron V1 Venti weights can be downloaded from the Hugging Face repository mindlab-research/Macaron-V1-Venti.
Basics
Macaron V1 Venti was made by Mind Lab and released on July 21, 2026.
Yes, Macaron V1 Venti is free to use because it has open weights.
Generated from Vertical Tabs data, updated November 11, 2026.
Similar models
Dots3-Note Preview
78.1BKolibri 1
120BNemotron 3.5 Super 120B A12B
Ornith 1.0 31BOrnith 1.0 31B
Ornith 1.0 35BOrnith 1.0 35B
Qwen3.8 Max Preview
36BApodex 1.1 Mini
Qwen3.8 Flash Next
ERNIE-5.5-Flash-PreviewERNIE-5.5-Flash-Preview
Claude Mythos 5.1
GPT-6 Astra (Fast)
MiniCPM-V 4.7 1B
Discussion
No comments yet
Be the first to share your thoughts.