DeepSeek-R1-0528 improves reasoning and logic via better computation and optimization, nearing the performance of top models like O3 and Gemini 2.5 Pro.

prompt

images

audios

videos

chat_history

audio

video

image

tools

tool_choice

The system-level prompt used to define the assistant's behavior.

system_prompt

The maximum number of tokens to generate. Shorter token lengths will provide faster performance.

max_tokens

A decimal number that determines the degree of randomness in the response.

temperature

An alternative to sampling with temperature, where the model considers the results of the tokens with top_p probability mass.

top_p

The level of reasoning effort to apply to the response. Currently supported values are low, medium, and high. 

reasoning_effort

predict

return

generate

The single model method to get the OpenAI-compatible request and send it to the OpenAI server
  then return its response.

Args:
    msg: JSON string containing the request parameters

Returns:
    JSON string containing the response or error

openai_transport

Process an OpenAI-compatible request and return a streaming response iterator.
This method is used when stream=True and returns an iterator of strings directly,
without converting to a list or JSON serializing.

Args:
    msg: The request as a JSON string.

Returns:
    Iterator[str]: An iterator yielding text chunks from the streaming response.

openai_stream_transport

Model is trained and ready for deployment

DeepSeek-R1-0528-Qwen3-8B

Llama 3.2 (3B) is a multilingual, instruction-tuned LLM by Meta, optimized for dialogue, retrieval, and summarization. It uses an autoregressive transformer with SFT and RLHF for improved alignment and outperforms many industry models.

Llama-3_2-3B-Instruct

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.

Qwen3-14B

Identifies a variety of concepts in images and video including objects, themes, and more. Trained with over 10,000 concepts and 20M images.

Image Recognition

Gemma-3n-E2B is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. Gemma 3n models are designed for efficient execution on low-resource devices

Process an OpenAI-compatible request and send it to the appropriate OpenAI endpoint.

Args:
    msg: JSON string containing the request parameters including 'openai_endpoint'

Returns:
    JSON string containing the response or error

Process an OpenAI-compatible request and return a streaming response iterator.
This method is used when stream=True and returns an iterator of strings directly,
without converting to a list or JSON serializing. Supports chat completions and responses endpoints.

Args:
    msg: The request as a JSON string.

Returns:
    Iterator[str]: An iterator yielding text chunks from the streaming response.

The system-level prompt that defines the assistant's behavior. If not provided, the model's default system prompt will be used.

gemma-3n-E2B-it-text

Gemma-3n-E4B is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. Gemma 3n models are designed for efficient execution on low-resource devices

gemma-3n-E4B-it-text

The MiniCPM4 series are efficient LLMs optimized for end-side devices, achieved through innovations in architecture, data, training, and inference.

MiniCPM4-8B

Grok-3 is a state-of-the-art large language model (LLM) developed by XAI, their most sophisticated model to date, which combines robust reasoning with vast pretraining knowledge.

The single model method to get the OpenAI-compatible request and send it to the OpenAI server
then return its response.

Args:
    msg: JSON string containing the request parameters

Returns:
    JSON string containing the response or error

This method is used to predict the output of the model using the grok-3 model with streaming = False.

An upper bound for the number of tokens that can be generated for a completion.

A decimal number that determines the degree of randomness in the response

This method is used to generate the output of the model using the grok-3 model with streaming support.

grok-3

Devstral is an agentic LLM for software engineering, developed by Mistral AI and All Hands AI. It’s designed to explore codebases, edit multiple files, and support engineering agents.

Devstral-Small-2505-unsloth-bnb-4bit

GPT-4o is a multimodal AI model that excels in processing and generating text, audio, and images, offering rapid response times and improved performance across languages and tasks, while incorporating advanced safety features

This method is used to predict the output of the model using the OpenAI gpt-4o model with streaming = False.

This method is used to generate the output of the model using the OpenAI gpt-4o model with streaming support.

gpt-4o

GPT-4.1 is OpenAI's advanced LLM, optimized for coding, instruction following, and processing extended contexts up to 1 million tokens. It delivers enhanced performance, making it ideal for complex and long-form content generation tasks

This method is used to predict the output of the model using the OpenAI gpt-4.1 model with streaming = False.

This method is used to generate the output of the model using the OpenAI gpt-4.1 model with streaming support.

gpt-4_1

Gemini 2.5 Flash-Lite is Google’s fastest, most efficient model yet, with improved performance and ultra-low latency.

This method is used to predict the output of the model using the Gemini model with streaming = False.

This method is used to predict the output of the model using the Gemini model with streaming = True.

gemini-2_5-flash-lite

Gemini 2.5 Flash Preview is the next iteration in the Gemini 2.0 series of models, a suite of highly-capable, natively multimodal, reasoning models. Gemini 2.5 Flash Preview is Google’s first fully hybrid reasoning model.

This method is used to predict the output of the model using the Gemini-2.5 PRO model with streaming = False.

This method is used to predict the output of the model using the Gemini-2.5 FLASH model with streaming = True.

gemini-2_5-flash

Gemini 2.0 Flash is a fast, low-latency multimodal model with enhanced performance and new capabilities

This method is used to predict the output of the model using the Gemini-2.0 FLASH model with streaming = False.

This method is used to predict the output of the model using the Gemini-2.0 FLASH model with streaming = True.

gemini-2_0-flash

Qwen3 is the newest in the Qwen series, featuring dense and MoE models with major improvements in reasoning, instruction-following, agent tasks, and multilingual support.

Qwen3-30B-A3B-FP8

Gemma 3 (12B) is a multilingual, multimodal open model by Google, handling text and image inputs with a 128K context window. It excels in tasks like QA and summarization while being efficient for deployment on limited-resource devices.

gemma-3-12b-it

Phi-4-reasoning-plus is a state-of-the-art open-weight reasoning model finetuned from Phi-4 using supervised fine-tuning on a dataset of chain-of-thought traces and reinforcement learning. 

Phi-4-reasoning-plus

MiniCPM3-4B is the 3rd generation of MiniCPM series. The overall performance of MiniCPM3-4B surpasses Phi-3.5-mini-Instruct and GPT-3.5-Turbo-0125, being comparable with many recent 7B~9B models.

MiniCPM3-4B

Phi-4-mini-instruct is a lightweight open model from the Phi-4 family, optimized for reasoning with high-quality data. It supports a 128K context window and uses fine-tuning for precise instruction adherence and safety.

Phi-4-mini-instruct

Qwen2.5-VL is a vision-language model designed for AI agents, finance, and commerce. It excels in visual recognition, reasoning, long video analysis, object localization, and structured data extraction.

Qwen2_5-VL-7B-Instruct

Phi-4 is a state-of-the-art open model trained on high-quality synthetic, public, and academic data for advanced reasoning. It uses fine-tuning and preference optimization for precise instruction adherence and safety.

phi-4

Grok-2-Vision-1212 is a cutting-edge multimodal AI model developed by xAI, integrating advanced language processing with sophisticated visual understanding.

This method is used to predict the output of the model using the grok-2-vision model with streaming = False.

This method is used to generate the output of the model using the grok-2-vision model with streaming support.

grok-2-vision-1212

The Grok-2 Image Generation Model (grok-2-image-1212) is xAI's advanced tool for transforming text prompts into high-quality images. Integrated with the Aurora autoregressive model, it delivers photorealistic visuals

This method is used to generate an image based on the provided text prompt using the grok-2-image model.
Args:
    prompt (str): The text prompt to generate the image from.
Returns:
    Image: An Image object

This method is used to generate multiple images based on the provided text prompt using the grok-2-image model.
Args:
    prompt (str): The text prompt to generate the image from.
    n (int): The number of images to generate. Default is 1.
Returns:
    List[Image]: A list of Image objects generated from the prompt.

generate_images

grok-2-image-1212

Grok-2 is xAI’s flagship large language model (LLM), designed to deliver advanced reasoning, multilingual understanding, and highly steerable outputs.

grok-2-1212

QwQ is the reasoning model of the Qwen series, designed for enhanced problem-solving and downstream task performance. QwQ-32B competes with top reasoning models like DeepSeek-R1 and o1-mini.

QwQ-32B-AWQ

Multi-model workflow that detects, crops, and recognizes demographic characteristics of faces. Visually classifies age, gender, and multi-culture characteristics.

Multi-model workflow that combines face detection and sentiment classification of 7 concepts: anger, disgust, fear, neutral, happiness, sadness, contempt, and surprise.

A general image workflow that combines detection, classification, and embedding to identify general concepts including objects, themes, moods,  etc.

RAG Agent uses  GPT-4 Turbo LLM model with ReAct prompting, optimizing dynamic reasoning and action planning.

Claude 3.5 Sonnet is a high-speed, advanced AI model excelling in reasoning, knowledge, coding, and visual tasks, ideal for complex applications.




claude-3_5-sonnet

DeepSeek-R1-Distill-Qwen-7B is a 7B-parameter dense model distilled from DeepSeek-R1 based on Qwen-7B.



DeepSeek-R1-Distill-Qwen-7B

Pixtral 12B is a natively multimodal model excelling in multimodal reasoning, instruction following, and text benchmarks with a 12B parameter architecture supporting variable image sizes and long context inputs

pixtral-12b

DeepSeek-R1-Distill-Qwen-32B is a 32B-parameter dense model distilled from DeepSeek-R1 based on Qwen-32B.



DeepSeek-R1-Distill-Qwen-32B

Llama 3.3 (70B) is a multilingual instruction-tuned LLM optimized for dialogue, trained on 15T+ tokens, supporting 8 languages, and incorporating strong safety measures

sglang-llama-3_3-70b-instruct

DeepSeek-Coder-33B-Instruct model is a SOTA 33 billion parameter code generation model, fine-tuned on 2 billion tokens of instruction data, offering superior performance in code completion and infilling tasks across more than 80 programming languages.

deepseek-coder-33b-instruct

MiniCPM-o is the latest series of end-side multimodal LLMs (MLLMs) ungraded from MiniCPM-V. The models can now take images, video, text, and audio as inputs and provide high-quality text and speech outputs in an end-to-end fashion

minicpm-o-2_6

DeepSeek-R1-Distill-Qwen-1_5B is a 1.5B-parameter dense model distilled from DeepSeek-R1 based on Qwen-1.5B.



DeepSeek-R1-Distill-Qwen-1_5B

DeepSeek-R1-Distill-Qwen-14B is a 14B-parameter dense model distilled from DeepSeek-R1 based on Qwen-14B.


DeepSeek-R1-Distill-Qwen-14B

Granite-3.0-8B-Instruct is a versatile, enterprise-ready language model optimized for multilingual understanding, coding, and instruction-following across diverse tasks and constrained environments.

granite-3_0-8b-instruct

Granite-3.0-2B-Instruct SLM is a lightweight, multilingual, enterprise-ready language model optimized for instruction-following and code understanding, ideal for versatile applications under Apache 2.0.

granite-3_0-2b-instruct

DBRX-Instruct is a state-of-the-art, efficient, open LLM by Databricks, capable of handling input length up to 32K tokens. The model excels at a broad set of natural language tasks such as: text summarization, question-answering, extraction, and coding.

dbrx-instruct

A powerful 7-billion parameter LLM trained on up to 8K sequence length with fine-tuning on instructional data, enabling robust long sequence modeling and various NLP applications.

xgen-7b-8k-inst-xparams

The Snowflake Arctic-Instruct Model is a cost-effective, enterprise-focused LLM that excels in SQL, coding, and instruction-following tasks, offering open access to its advanced AI capabilities

snowflake-arctic-instruct

openchat-3.6-8b model is a high-performance, open-source LLM fine-tuned from llama3-8b with C-RLFT, delivering ChatGPT-level performance across various tasks.

openchat-3_6-8b-20240522

OpenChat-3.5 is a versatile 7B LLM, fine-tuned with C-RLFT, excelling in benchmarks with competitive scores, supporting diverse use cases from general chat to mathematical problem-solving.

openchat-3_5-1210

Genstruct 7B is an instruction-generation model (llm), designed to create valid instructions given a raw text corpus. It enables the creation of new, partially synthetic instruction finetuning datasets from any raw-text corpus.


genstruct-7b


Florence-2-large is a lightweight, versatile vision-language model by Microsoft, excelling in multiple tasks using a unified representation and the extensive FLD-5B dataset

florence-2-large

phi-1 is a high-performing 1.3 billion-parameter text-to-code language model, excelling in python code generation tasks while prioritizing high-quality training data

phi-1-alternative

Phi-1.5 is a 1.3 billion parameter language model(llm) that excels at complex reasoning tasks and was trained on a high-quality synthetic dataset.

phi-1_5

Phi-2 is a 2.7 billion-parameter LLM achieving state-of-the-art performance in QA, chat, and code tasks, with a focus on high-quality training data, demonstrating improved behavior in toxicity and bias.




phi-2

The Phi-3-Vision-128K-Instruct is a high-performance, cost-effective multimodal model for advanced text and image understanding tasks.

phi-3-vision-128k-instruct

Phi-3-Mini-4K-Instruct is a 3.8B parameter small language model offering state-of-the-art performance in reasoning and instruction-following tasks, outperforming larger models with its high-quality data training.




phi-3-mini-4k

MiniCPM-Llama3-V 2.5 is a high-performance, efficient 8B parameter  multimodal model excelling in OCR, multilingual support, and multimodal tasks.




miniCPM-Llama3-V-2_5

Gemini 2.0 Flash-Lite is our fastest and most cost efficient Flash model. It's an upgrade path for 1.5 Flash users who want better quality for the same price and speed.

This method is used to predict the output of the model using the Gemini-2.0 FLASH LITE model with streaming = False.

This method is used to predict the output of the model using the Gemini-2.0 FLASH LITE model with streaming = True.

gemini-2_0-flash-lite

LLaVA-1.5 is a state-of-the-art language vision model that represents a significant advancement in the field of multimodal artificial intelligence

llava-1_5-7b

Gemini 1.5 Pro is a powerful, efficient AI LLM with 1 million long-context window, enabling advanced reasoning and comprehension across various data types. 


gemini-1_5-pro

Gemini 1.5 Flash is a cost-efficient, high-speed foundation llm optimized for multimodal tasks, ideal for applications requiring rapid processing and scalability

gemini-1_5-flash

The Claude 3 Opus model is a state-of-the-art, multimodal language model (llm) with superior performance in reasoning, math, coding, and multilingual understanding.

claude-3-opus

Claude 3 Sonnet is a multimodal AI llm model balancing skills and speed, excelling in reasoning, multilingual tasks, and visual interpretation.

claude-3-sonnet

GPT-4o Mini: An affordable, high-performing small model excelling in text and vision tasks with extensive context support

gpt-4o-mini

CogVLM-Chat is a state-of-the-art visual language model that excels in generating context-aware, conversational responses by integrating advanced visual and textual understanding

cogvlm-chat

Qwen-VL-Chat is a high-performing LVLM by Alibaba Cloud for text-image dialogue tasks, excelling in zero-shot captioning, VQA, and referring expression comprehension while supporting multilingual dialogue.






qwen-VL-Chat

Fuyu-8B is an open-source, simplified multimodal architecture with a decoder-only transformer, supporting arbitrary image resolutions, and excelling in diverse applications, including question answering and complex visual understanding

fuyu-8b

Explore Clarifai's free community for AI resources. Use our state-of-the-art AI models, workflows and more to add AI into your own applications.