Xiaomi releases the open-source large model MiMo-V2-Flash, which is designed for high speed and efficiency, showing outstanding performance in tasks such as inference and code generation, with response speed surpassing multiple popular domestic models. The model adopts a sparse activation architecture, with 309 billion parameters, and the weights and code are open-sourced under the MIT license.
Research indicates that multi-agent systems exhibit significant performance fluctuations, with task type being a key factor. Centralized architectures perform better in parallel tasks.....
MIT researchers developed an instance-adaptive scaling technique that dynamically adjusts computational resources for large language models based on problem complexity, enhancing efficiency and reducing energy consumption. Supported by multiple institutions, the related paper was released in early November.....
MIT scientists created the first millimeter-scale probability map of the human brain's language network after 15 years of research. This compact yet powerful network, distinct from thought and emotion modules, specializes in word meaning mapping and sentence assembly, confirmed through 1,400 fMRI scans.....
Enzzo AI is an AI-driven PRD solution that compresses product requirement documents, generates requirements, mitigates risks, promotes team collaboration, and improves efficiency.
A big data analysis solution based on data warehouses that helps you make decisions more quickly and efficiently, driving business growth.
Master your health, predict your future, and transform your health trajectory.
Automated interpretability agent enhancing AI model transparency
Anthropic
$21
Input tokens/M
$105
Output tokens/M
200
Context Length
Alibaba
$3.9
$15.2
64
-
Deepseek
$4
$12
128
Chatglm
$16
32
$1.8
$5.4
16
$0.7
$1.95
$2
$8
Minimax
Baidu
$0.8
$3.2
Stepfun
$0.1
$0.4
mradermacher
This project provides a static quantized version of the Qwen-4B-Instruct-2507-Self-correct model, supporting tasks such as text generation, bias mitigation, and self-correction. Based on the Qwen-4B architecture, the model has undergone instruction fine-tuning and self-correction training, offering multiple quantized versions to meet different hardware requirements.
mitegvg
This model is a violence detection model based on the VideoMAE architecture. After pre-training on the Kinetics dataset, it was fine-tuned for 92 rounds for the violence detection task. The model uses the Vision Transformer architecture and is specifically designed for video content analysis, capable of identifying violent behaviors in videos.
mlx-community
MiniMax-M2-5bit is a 5-bit quantized version converted from the MiniMaxAI/MiniMax-M2 model, optimized specifically for the MLX framework. This model is a large language model that supports text generation tasks and is released under the MIT license.
Mitchins
This is a deep learning model based on the EfficientNet-B0 architecture, specifically designed for classifying the art styles of anime and visual novel images. The model can accurately identify 6 different anime art styles, including dark, flat, modern, cute, painting, and retro styles.
jeevanrushi07
OpenLLaMA 3B v2 is an open-source large language model based on the Transformer architecture, with 3 billion parameters. This model uses the MIT license and is mainly used for English text generation tasks, supporting various application scenarios such as chatbots.
TareksLab
This is a large language model with 70B parameters merged using the Linear DELLA method. Based on the Llama-3.1-Nemotron-lorablated-70B base model, it integrates multiple high-performance Llama-3.3-70B variant models, aiming to provide more powerful language understanding and generation capabilities.
cpatonn
GLM-4.5-Air-AWQ is an 8-bit quantized version based on the GLM-4.5-Air base model, designed specifically for intelligent agents. It uses a hybrid reasoning mode, supports complex reasoning and instant response, and is released under the MIT open-source license.
unsloth
GLM-4.5 is a foundation model designed for intelligent agents, integrating inference, coding, and intelligent agent capabilities. It has a total of 355 billion parameters and ranks 3rd in 12 industry-standard benchmark evaluations, scoring 63.2. It uses the MIT open-source license and can be used for commercial purposes and secondary development.
This is a static quantized version of the FBogaerts/NextCoder-7B-Finetuned model, specifically optimized for code generation and text generation tasks, providing a more efficient inference solution. The model supports English and is released under the MIT license.
GLM-4.5-Air-AWQ is a text generation model that performs 4-bit AWQ quantization based on the zai-org/GLM-4.5-Air base model. It is specifically designed for intelligent agent applications and performs excellently in reasoning, coding, and intelligent agent capabilities. It uses the MIT open-source license.
lmstudio-community
GLM-4.5-Air is an efficient language model developed by zai-org. It has been optimized by the MLX team with 8-bit quantization and specifically optimized for Apple Silicon chips. This model is based on the transformers architecture, supports text generation tasks, and is released under the MIT license.
HugoHE
M-Hood is a series of models specifically designed to mitigate the hallucination phenomenon in object detection. Through novel fine-tuning strategies and a revised benchmark dataset, it significantly reduces false alarms on out-of-distribution data and enhances the safety and reliability of object detection systems.
mit-han-lab
The FLUX.1-Kontext-dev version processed by Nunchaku quantization can edit images according to text instructions, optimizing inference efficiency while minimizing performance loss.
A Nunchaku quantized version based on Shuttle Jaguar, designed to generate high-quality images according to text prompts, optimizing efficient inference and minimizing performance loss.
The Nunchaku quantized version of FLUX.1-Fill-dev, capable of filling regions in existing images based on text descriptions, achieving efficient inference through quantization processing.
The Nunchaku-quantized FLUX.1-schnell model is used to generate high-quality images based on text prompts and has been optimized for efficient inference.
The quantized version of FLUX.1-Depth-dev provided by the Nunchaku team can generate images based on text descriptions while preserving the structure of the input image, optimizing the inference efficiency.
The Nunchaku quantized version of FLUX.1-Canny-dev, capable of generating images based on text descriptions while following the structure of the given input image.
NICOPOI-9
An image segmentation model fine-tuned based on nvidia/mit-b3, performing excellently on the NICOPOI-9/morphpad_coord_hgo_512_4class_v2 dataset
SegFormer image segmentation model based on NVIDIA MIT-B1 architecture, fine-tuned on specific datasets, excels in high-precision image segmentation tasks
A server project implementing the MCP protocol, providing note storage and summary generation functions
An ATT&CK knowledge base query service based on the MCP protocol, providing functions for retrieving attack techniques and tactics, querying mitigation measures, and detection methods.
The Devici MCP Server is a model context protocol server used to interact with the Devici API, providing LLM tools to manage users, collections, threat models, components, threats, mitigation measures, teams, and dashboard data.
This project is a collaborative development project of Vale-MCP, mainly used for text verification and writing assistance, and adopts the MIT open-source license.
This project contains the specifications and protocol patterns of the model context protocol, provides definitions in both TypeScript and JSON Schema formats, and includes an open - source contribution guide and MIT license.
The project uses the MIT open-source license
This project implements an MCP server that interacts with Miden developer tools, provides developer documentation search functions, is built based on the MCP SDK, and uses the MIT license.
This project is a practice project for the MIT Hack decentralized AI summit, demonstrating how to quickly set up a Claude MCP protocol service on Windows. It includes two examples: a weather application and an SQLite database, which respectively demonstrate government API data acquisition and bakery inventory management functions.
Firefox MCP Bridge is a server-side program for the Claude web application, providing support for the配套 Firefox extension and licensed under both the MIT and Apache licenses.
An MCP server project developed by Yusuke Wada, which uses Node.js to run the game server and adopts the MIT license
The MCP server is used to access the MITRE ATTACK knowledge base
%20--%3e%3cdefs%3e%3cstyle%3e%20.st0%20{%20fill:%20%23061b40;%20}%20.st1%20{%20fill:%20%23306af1;%20}%20.st2%20{%20fill:%20%235ce5cf;%20}%20%3c/style%3e%3c/defs%3e%3cg%3e%3cpath%20class='st0'%20d='M55,10.5h9v3h-9v9h12V7.5h-12v3ZM64,19.5h-6v-3h6v3Z'/%3e%3cpolygon%20class='st0'%20points='69%2016.5%2078%2016.5%2078%2019.5%2069%2019.5%2069%2022.5%2081%2022.5%2081%2013.5%2072%2013.5%2072%2010.5%2081%2010.5%2081%207.5%2069%207.5%2069%2016.5'/%3e%3cpolygon%20class='st0'%20points='95%2010.5%2095%207.5%2083%207.5%2083%2022.5%2095%2022.5%2095%2019.5%2086%2019.5%2086%2016.5%2095%2016.5%2095%2013.5%2086%2013.5%2086%2010.5%2095%2010.5'/%3e%3cpath%20class='st0'%20d='M40,1.5v21h11.6l1.4-1.4v-7.6h0c0,0-1.4-1.5-1.4-1.5l1.4-1.4V2.9l-1.4-1.4h-11.6ZM50,19.5h-7v-6h7v6ZM50,10.5h-7v-6h7v6Z'/%3e%3c/g%3e%3cpath%20class='st1'%20d='M23.1,24L14.7,4.8l-4.9,11.2h3.8l-1.8,4H3.3L12.1,0H2C.9,0,0,.9,0,2v20c0,1.1.9,2,2,2h21.1Z'/%3e%3cpath%20class='st2'%20d='M34,0h-16.8l10.6,24h6.2c1.1,0,2-.9,2-2V2C36,.9,35.1,0,34,0ZM32.5,20h-4V4h4v16Z'/%3e%3c/svg%3e)
