浏览 AI 模型

Nous Research

Dolphin 13B is a general-purpose uncensored model fine-tuned for broad capabilities including coding, reasoning, and creative writing without alignment restrictions.

Nous Research

Nous Hermes is a fine-tuned model optimized for instruction following and helpful dialogue. Trained on curated datasets emphasizing quality responses, reasoning, and user alignment.

Allen AI

OLMo 2 7B is Allen AI's fully open language model with open data, code, and weights.

Instinct AI

Solar 7B is Upstage's efficient language model built on a depth-upscaled architecture. Offers strong instruction following and reasoning performance optimized for single-GPU inference.

LMSYS

Vicuna is a chat assistant trained by fine-tuning Llama 2 on user-shared conversations collected from ShareGPT.

WizardLM

Project Repo: https://github.com/nlpxucan/WizardLM

WizardLM

📃 [WizardLM] • 📃 [WizardCoder] • 📃 [WizardMath]

Microsoft

Phi-4-mini-instruct is a lightweight open model built upon synthetic data and filtered publicly available websites - with a focus on high-quality, reasoning dense data. The model belongs to the Phi-4 model family and supports 128K token context length. The model underwent an enhancement process, incorporating both supervised fine-tuning and direct preference optimization to support precise instruction adherence and robust safety measures.

Alibaba

Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5:

Alibaba

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features:

Google

Gemma 2 27B is Google's largest Gemma 2 model, offering state-of-the-art performance among open models of similar size. Built on Gemini technology with strong reasoning, code, and multilingual capabilities.

Alibaba

Qwen 2.5 3B provides a good balance of capability and efficiency, suitable for laptops and entry-level GPUs.

Alibaba

Qwen 2.5 Coder 1.5B is Alibaba's compact code-specific language model from the Qwen2.5 Coder series. Trained on 5.5T tokens including source code, text-code grounding, and synthetic data. Features improvements in code generation, reasoning, and fixing while maintaining general and math capabilities.

LLaVA

Model type: LLaVA is an open-source chatbot trained by fine-tuning LLaMA/Vicuna on GPT-generated multimodal instruction-following data. It is an auto-regressive language model, based on the transformer architecture.

Cohere

Command R+ is Cohere's most capable open-weight model for enterprise RAG workloads. Offers superior long-context reasoning, multi-step tool use, and grounded generation with citations across 10 languages.

IBM

Granite 4.1 3B is IBM's smallest Granite 4.1 dense decoder-only model, trained on roughly 15T tokens with 128K context. Apache 2.0 licensed and tuned for fast, commercially-friendly RAG, coding, and assistant workloads on small GPUs.

Microsoft

The Phi-3-Mini-128K-Instruct is a 3.8 billion-parameter, lightweight, state-of-the-art open model trained using the Phi-3 datasets. This dataset includes both synthetic data and filtered publicly available website data, with an emphasis on high-quality and reasoning-dense properties. The model belongs to the Phi-3 family with the Mini version in two variants 4K and 128K which is the context length (in tokens) that it can support.

DeepSeek

DeepSeek R1 Distill Qwen 7B is a 7B-parameter reasoning model distilled from the larger DeepSeek-R1. Based on Qwen2.5-Math-7B and fine-tuned on 800K samples from DeepSeek-R1, it delivers strong reasoning with 92.8% on MATH-500 and 49.1 on GPQA Diamond while being far more efficient than the full 671B model.

DeepSeek

We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrated remarkable performance on reasoning. With RL, DeepSeek-R1-Zero naturally emerged with numerous powerful and interesting reasoning behaviors. However, DeepSeek-R1-Zero encounters challenges such as endless repetition, poor readability, and language mixing.

Baichuan

Baichuan-13B-Chat为Baichuan-13B系列模型中对齐后的版本，预训练模型可见Baichuan-13B-Base。

Baichuan

Baichuan-7B是由百川智能开发的一个开源的大规模预训练模型。基于Transformer结构，在大约1.2万亿tokens上训练的70亿参数模型，支持中英双语，上下文窗口长度为4096。在标准的中文和英文权威benchmark（C-EVAL/MMLU）上均取得同尺寸最好的效果。

Cerebras

Check out our Blog Post and arXiv paper!

TII

Falcon-7B-Instruct is a 7B parameters causal decoder-only model built by TII based on Falcon-7B and finetuned on a mixture of chat/instruct datasets. It is made available under the Apache 2.0 license.

IBM

Granite Code 3B is IBM's compact code generation model for enterprise use.