o1-mini is a fast and cost-effective reasoning model designed for programming, mathematics, and scientific applications. This model features a 128K context and has a knowledge cutoff date of October 2023.

o1 is OpenAI's new reasoning model, suitable for complex tasks that require extensive general knowledge. This model features a 128K context and has a knowledge cutoff date of October 2023.

ChatGPT-4o is a dynamic model that updates in real-time to stay current with the latest version. It combines powerful language understanding and generation capabilities, making it suitable for large-scale applications, including customer service, education, and technical support.

GPT-4o mini is the latest model released by OpenAI after GPT-4 Omni, supporting both image and text input while outputting text. As their most advanced small model, it is significantly cheaper than other recent cutting-edge models, costing over 60% less than GPT-3.5 Turbo. It maintains state-of-the-art intelligence while offering remarkable cost-effectiveness. GPT-4o mini scored 82% on the MMLU test and currently ranks higher than GPT-4 in chat preferences.

Gemini 1.5 Pro supports up to 2 million tokens, making it an ideal choice for medium-sized multimodal models, providing multifaceted support for complex tasks.

Gemini 1.5 Flash is Google's latest multimodal AI model, featuring fast processing capabilities and supporting text, image, and video inputs, making it suitable for efficient scaling across various tasks.

Claude 3.5 Sonnet offers capabilities that surpass Opus and faster speeds than Sonnet, while maintaining the same price as Sonnet. Sonnet excels particularly in programming, data science, visual processing, and agent tasks.

Claude 3 Haiku is Anthropic's fastest and most compact model, designed for near-instantaneous responses. It features rapid and accurate directional performance.

Claude 3 Opus is Anthropic's most powerful model for handling highly complex tasks. It excels in performance, intelligence, fluency, and comprehension.

Mistral Nemo is a 12B model developed in collaboration with Nvidia, offering outstanding reasoning and coding performance, easy to integrate and replace.

Mistral Small is a cost-effective, fast, and reliable option suitable for use cases such as translation, summarization, and sentiment analysis.

Mistral Large is the flagship model, excelling in multilingual tasks, complex reasoning, and code generation, making it an ideal choice for high-end applications.

Codestral is a cutting-edge generative model focused on code generation, optimized for intermediate filling and code completion tasks.

The Pixtral model demonstrates strong capabilities in tasks such as chart and image understanding, document question answering, multimodal reasoning, and instruction following. It can ingest images at natural resolutions and aspect ratios and handle an arbitrary number of images within a long context window of up to 128K tokens.

Llama 3.1 Sonar Small Online model, featuring 8B parameters, supports a context length of approximately 127,000 tokens, designed for online chat, efficiently handling various text interactions.

Llama 3.1 Sonar Large Online model, featuring 70B parameters, supports a context length of approximately 127,000 tokens, suitable for high-capacity and diverse chat tasks.

Llama 3.1 Sonar Huge Online model, featuring 405B parameters, supports a context length of approximately 127,000 tokens, designed for complex online chat applications.

A new open-source model that integrates general and coding capabilities, retaining the general conversational abilities of the original Chat model and the powerful code handling capabilities of the Coder model, while better aligning with human preferences. Additionally, DeepSeek-V2.5 has achieved significant improvements in writing tasks, instruction following, and more.

R1 is the reasoning model launched by DeepSeek. Before outputting the final answer, the model first provides a chain of thought to enhance the accuracy of the final response.

o4-mini.description

o3.description

gpt-4.1-mini.description

gpt-4.1-nano.description

ChatGPT-4o is a dynamic model that updates in real-time to maintain the latest version. It combines powerful language understanding and generation capabilities, making it suitable for large-scale applications including customer service, education, and technical support.

gpt-4.1.description

ChatGPT-4o is a dynamic model that updates in real-time to stay current with the latest version. It combines powerful language understanding and generation capabilities, making it suitable for large-scale applications, including customer service, education, and technical support.

ChatGPT-4o is a dynamic model that updates in real-time to stay current with the latest version. It combines powerful language understanding and generation capabilities, making it suitable for large-scale applications, including customer service, education, and technical support.

ChatGPT-4o is a dynamic model that updates in real-time to stay current with the latest version. It combines powerful language understanding and generation capabilities, making it suitable for large-scale applications, including customer service, education, and technical support.

The latest GPT-4 Turbo model features visual capabilities. Now, visual requests can be made using JSON format and function calls. GPT-4 Turbo is an enhanced version that provides cost-effective support for multimodal tasks. It strikes a balance between accuracy and efficiency, making it suitable for applications requiring real-time interaction.

The latest GPT-4 Turbo model features visual capabilities. Now, visual requests can be made using JSON format and function calls. GPT-4 Turbo is an enhanced version that provides cost-effective support for multimodal tasks. It strikes a balance between accuracy and efficiency, making it suitable for applications requiring real-time interaction.

The latest GPT-4 Turbo model features visual capabilities. Now, visual requests can be made using JSON format and function calls. GPT-4 Turbo is an enhanced version that provides cost-effective support for multimodal tasks. It strikes a balance between accuracy and efficiency, making it suitable for applications requiring real-time interaction.

The latest GPT-4 Turbo model features visual capabilities. Now, visual requests can be made using JSON format and function calls. GPT-4 Turbo is an enhanced version that provides cost-effective support for multimodal tasks. It strikes a balance between accuracy and efficiency, making it suitable for applications requiring real-time interaction.

The latest GPT-4 Turbo model features visual capabilities. Now, visual requests can be made using JSON format and function calls. GPT-4 Turbo is an enhanced version that provides cost-effective support for multimodal tasks. It strikes a balance between accuracy and efficiency, making it suitable for applications requiring real-time interaction.

GPT-4 offers a larger context window, capable of handling longer text inputs, making it suitable for scenarios that require extensive information integration and data analysis.

GPT-4 offers a larger context window, capable of handling longer text inputs, making it suitable for scenarios that require extensive information integration and data analysis.

GPT-4 offers a larger context window, capable of handling longer text inputs, making it suitable for scenarios that require extensive information integration and data analysis.

GPT-4 offers a larger context window, capable of handling longer text inputs, making it suitable for scenarios that require extensive information integration and data analysis.

GPT 3.5 Turbo is suitable for various text generation and understanding tasks. Currently points to gpt-3.5-turbo-0125.

GPT 3.5 Turbo is suitable for various text generation and understanding tasks. Currently points to gpt-3.5-turbo-0125.

GPT 3.5 Turbo is suitable for various text generation and understanding tasks. Currently points to gpt-3.5-turbo-0125.

GPT 3.5 Turbo is suitable for various text generation and understanding tasks. Currently points to gpt-3.5-turbo-0125.

Llama 3.1 is a leading model launched by Meta, supporting up to 405B parameters, applicable in complex dialogues, multilingual translation, and data analysis.

70b.description

405b.description

Code Llama is an LLM focused on code generation and discussion, combining extensive programming language support, suitable for developer environments.

13b.description

34b.description

70b.description

QwQ is an experimental research model focused on improving AI reasoning capabilities.

0.5b.description

1.5b.description

qwen2.5.description

72b.description

CodeQwen1.5 is a large language model trained on extensive code data, specifically designed to solve complex programming tasks.

0.5b.description

1.5b.description

Qwen2 is Alibaba's next-generation large-scale language model, supporting diverse application needs with excellent performance.

72b.description

2b.description

Gemma 2 is an efficient model launched by Google, covering a variety of application scenarios from small applications to complex data processing.

27b.description

2b.description

CodeGemma is a lightweight language model dedicated to various programming tasks, supporting rapid iteration and integration.

Phi-3 is a lightweight open model launched by Microsoft, suitable for efficient integration and large-scale knowledge reasoning.

14b.description

WizardLM 2 is a language model provided by Microsoft AI, excelling in complex dialogues, multilingual capabilities, reasoning, and intelligent assistant applications.

8x22b.description

MathΣtral is designed for scientific research and mathematical reasoning, providing effective computational capabilities and result interpretation.

Mistral is a 7B model released by Mistral AI, suitable for diverse language processing needs.

Mixtral is an expert model from Mistral AI, featuring open-source weights and providing support in code generation and language understanding.

8x22b.description

Mixtral Large is Mistral's flagship model, combining capabilities in code generation, mathematics, and reasoning, supporting a 128k context window.

Mistral Nemo, developed in collaboration with Mistral AI and NVIDIA, is a high-performance 12B model.

Codestral is Mistral AI's first code model, providing excellent support for code generation tasks.

Aya 23 is a multilingual model launched by Cohere, supporting 23 languages, facilitating diverse language applications.

35b.description

Command R is an LLM optimized for dialogue and long context tasks, particularly suitable for dynamic interactions and knowledge management.

Command R+ is a high-performance large language model designed for real enterprise scenarios and complex applications.

DeepSeek V2 is an efficient Mixture-of-Experts language model, suitable for cost-effective processing needs.

236b.description

DeepSeek Coder V2 is an open-source hybrid expert code model that performs excellently in coding tasks, comparable to GPT4-Turbo.

236b.description

LLaVA is a multimodal model that combines a visual encoder with Vicuna for powerful visual and language understanding.

13b.description

34b.description

MiniCPM-V is a next-generation multimodal large model launched by OpenBMB, boasting exceptional OCR recognition and multimodal understanding capabilities, supporting a wide range of application scenarios.

0.description

Claude 3.5 Haiku is Anthropic's fastest next-generation model. Compared to Claude 3 Haiku, Claude 3.5 Haiku has improved in various skills and has surpassed the previous generation's largest model, Claude 3 Opus, in many intelligence benchmark tests.

0.description

Claude 3 Sonnet provides an ideal balance of intelligence and speed for enterprise workloads. It offers maximum utility at a lower price, reliable and suitable for large-scale deployment.

Claude 2 provides advancements in key capabilities for enterprises, including industry-leading 200K token context, significantly reducing the occurrence of model hallucinations, system prompts, and a new testing feature: tool invocation.

Claude 2 provides advancements in key capabilities for enterprises, including industry-leading 200K token context, significantly reducing the occurrence of model hallucinations, system prompts, and a new testing feature: tool invocation.

0.description

0.description

0.description

0.description

0.description

0.description

0.description

0.description

1.description

Anthropic's model demonstrates high capability across a wide range of tasks, from complex conversations and creative content generation to detailed instruction following.

A fast, economical, yet still highly capable model that can handle a range of tasks, including everyday conversations, text analysis, summarization, and document Q&A.

0.description

0.description

0.description

0.description

0.description

gemini-2.5-pro-preview-03-25.description

gemini-2.0-flash-thinking-exp-01-21.description

gemini-2.0-flash-thinking-exp-1219.description

gemini-2.0-flash.description

gemini-2.5-flash-standard.description

gemini-2.5-flash-thinking.description

gemini-exp-1206.description

gemini-exp-1121.description

Gemini Exp 1114 is Google's latest experimental multimodal AI model, featuring rapid processing capabilities and supporting text, image, and video inputs, making it suitable for efficient scaling across various tasks.

learnlm-1.5-pro-experimental.description

Gemini 1.5 Flash 002 is an efficient multimodal model that supports a wide range of applications.

Gemini 1.5 Flash 001 is an efficient multimodal model that supports extensive application scaling.

Gemini 1.5 Pro 002 is the latest production-ready model, delivering higher quality outputs, with notable enhancements in mathematics, long-context, and visual tasks.

Gemini 1.5 Pro 001 is a scalable multimodal AI solution that supports a wide range of complex tasks.

Gemini 1.5 Flash 8B is an efficient multimodal model that supports a wide range of applications.

Gemini 1.5 Flash 8B 0924 is the latest experimental model, showcasing significant performance improvements in both text and multimodal use cases.

Gemini 1.0 Pro is Google's high-performance AI model, designed for extensive task scaling.

Gemini 1.0 Pro 001 (Tuning) offers stable and tunable performance, making it an ideal choice for complex task solutions.

Gemini 1.0 Pro 002 (Tuning) provides excellent multimodal support, focusing on effective solutions for complex tasks.

Mistral (7B) Instruct v0.3 offers efficient computational power and natural language understanding, suitable for a wide range of applications.

Google's lightweight instruction-tuning model.

A large language model developed by the Alibaba Cloud Tongyi Qianwen team

Qwen2.5-Coder focuses on code writing.

QwQ-32B-Preview is Qwen's latest experimental research model, focusing on enhancing AI reasoning capabilities. By exploring complex mechanisms such as language mixing and recursive reasoning, its main advantages include strong analytical reasoning, mathematical, and programming abilities. However, it also faces challenges such as language switching issues, reasoning loops, safety considerations, and differences in other capabilities.

microsoft/Phi-3.5-mini-instruct.description

NousResearch/Hermes-3-Llama-3.1-8B.description

deepseek-ai/DeepSeek-R1-Distill-Qwen-32B.description

Based on context length, topic, and complexity, your request will be sent to Llama 3 70B Instruct, Claude 3.5 Sonnet (self-regulating), or GPT-4o.

openai/o1.description

Claude 3 Haiku is Anthropic's fastest and most compact model, designed for near-instantaneous responses. It features quick and accurate directional performance.

Claude 3.5 Sonnet offers capabilities that surpass Opus and faster speeds than Sonnet, while maintaining the same pricing as Sonnet. Sonnet excels particularly in programming, data science, visual processing, and agent tasks.

Claude 3 Opus is Anthropic's most powerful model for handling highly complex tasks. It excels in performance, intelligence, fluency, and comprehension.

Gemini 1.5 Flash offers optimized multimodal processing capabilities, suitable for various complex task scenarios.

Gemini 1.5 Pro combines the latest optimization technologies to deliver more efficient multimodal data processing capabilities.

LLaMA 3.2 is designed to handle tasks that combine visual and textual data. It excels in tasks such as image description and visual question answering, bridging the gap between language generation and visual reasoning.

LLaMA 3.2 is designed to handle tasks that combine visual and textual data. It excels in tasks such as image description and visual question answering, bridging the gap between language generation and visual reasoning.

free.description

free.description

free.description

@hf/thebloke/deepseek-coder-6.7b-instruct-awq.description

@hf/google/gemma-7b-it.description

@hf/nousresearch/hermes-2-pro-mistral-7b.description

@cf/meta/llama-3-8b-instruct-awq.description

@cf/openchat/openchat-3.5-0106.description

@cf/qwen/qwen1.5-14b-chat-awq.description

@hf/nexusflow/starling-lm-7b-beta.description

Generation over generation, Meta Llama 3 demonstrates state-of-the-art performance on a wide range of industry benchmarks and offers new capabilities, including improved reasoning.

DeepSeek-R1.description

A 52B parameter (12B active) multilingual model, offering a 256K long context window, function calling, structured output, and grounded generation.

A 398B parameter (94B active) multilingual model, offering a 256K long context window, function calling, structured output, and grounded generation.

Command R is a scalable generative model targeting RAG and Tool Use to enable production-scale AI for enterprises.

Command R+ is a state-of-the-art RAG-optimized model designed to tackle enterprise-grade workloads.

Mistral Small can be used for any language-based task that requires high efficiency and low latency.

Codestral-2501.description

The Llama 3.1 instruction-tuned text-only models are optimized for multilingual dialogue use cases and outperform many of the available open-source and closed chat models on common industry benchmarks.

The Llama 3.1 instruction-tuned text-only models are optimized for multilingual dialogue use cases and outperform many of the available open-source and closed chat models on common industry benchmarks.

The Llama 3.1 instruction-tuned text-only models are optimized for multilingual dialogue use cases and outperform many of the available open-source and closed chat models on common industry benchmarks.

A versatile 8-billion parameter model optimized for dialogue and text generation tasks.

A powerful 70-billion parameter model excelling in reasoning, coding, and broad language applications.

Phi-4.description

Phi-3.5-MoE-instruct.description

An updated version of the Phi-3-vision model.

The smallest member of the Phi-3 family, optimized for both quality and low latency.

The same Phi-3-mini model, but with a larger context size for RAG or few-shot prompting.

A 7B parameter model that provides better quality than Phi-3-mini, focusing on high-quality, reasoning-dense data.

The same Phi-3-small model, but with a larger context size for RAG or few-shot prompting.

A 14B parameter model that provides better quality than Phi-3-mini, focusing on high-quality, reasoning-dense data.

The same Phi-3-medium model, but with a larger context size for RAG or few-shot prompting.

Llama 3.1 8B Instruct is the latest version released by Meta, optimized for high-quality conversational scenarios, outperforming many leading closed-source models.

Llama 3.1 70B Instruct is designed for high-quality conversations, excelling in human evaluations, particularly in highly interactive scenarios.

Llama 3.1 405B Instruct is the latest version from Meta, optimized for generating high-quality dialogues, surpassing many leading closed-source models.

Llama 3 8B Instruct is optimized for high-quality conversational scenarios, performing better than many closed-source models.

Llama 3 70B Instruct is optimized for high-quality conversational scenarios, demonstrating excellent performance in various human evaluations.

Gemma 2 is Google's lightweight open-source text model series.

Mistral 7B Instruct is a high-performance industry-standard model optimized for speed and long context support.

WizardLM 2 7B is Microsoft's latest lightweight AI model, performing nearly ten times better than existing leading open-source models.

WizardLM-2 8x22B is Microsoft's state-of-the-art Wizard model, demonstrating extremely competitive performance.

Dolphin Mixtral 8x22B is a model designed for instruction following, dialogue, and programming.

Hermes 2 Pro Llama 3 8B is an upgraded version of Nous Hermes 2, featuring the latest internally developed datasets.

Hermes 2 Mixtral 8x7B DPO is a highly flexible multi-model fusion designed to provide an exceptional creative experience.

MythoMax l2 13B is a language model that combines creativity and intelligence by merging multiple top models.

OpenChat 7B is an open-source language model library fine-tuned using the 'C-RLFT (Conditional Reinforcement Learning Fine-Tuning)' strategy.

LLaMA 3.2 is designed for tasks involving both visual and textual data. It excels in tasks like image description and visual question answering, bridging the gap between language generation and visual reasoning.

LLaMA 3.2 is designed for tasks involving both visual and textual data. It excels in tasks like image description and visual question answering, bridging the gap between language generation and visual reasoning.

LLaMA 3.2 is designed for tasks involving both visual and textual data. It excels in tasks like image description and visual question answering, bridging the gap between language generation and visual reasoning.

LLaMA 3.2 is designed for tasks involving both visual and textual data. It excels in tasks like image description and visual question answering, bridging the gap between language generation and visual reasoning.

Llama 3.1 8B model utilizes FP8 quantization, supporting up to 131,072 context tokens, making it a standout in open-source models, excelling in complex tasks and outperforming many industry benchmarks.

Llama 3.1 70B model is finely tuned for high-load applications, quantized to FP8 for enhanced computational efficiency and accuracy, ensuring outstanding performance in complex scenarios.

The 405B Llama 3.1 Turbo model provides massive context support for big data processing, excelling in large-scale AI applications.

Llama 3.1 Nemotron 70B is a large language model customized by NVIDIA, designed to enhance the helpfulness of LLM-generated responses to user queries. The model has excelled in benchmark tests such as Arena Hard, AlpacaEval 2 LC, and GPT-4-Turbo MT-Bench, ranking first in all three automatic alignment benchmarks as of October 1, 2024. The model is trained using RLHF (specifically REINFORCE), Llama-3.1-Nemotron-70B-Reward, and HelpSteer2-Preference prompts based on the Llama-3.1-70B-Instruct model.

Llama 3 8B Instruct Turbo is a high-performance large language model, supporting a wide range of application scenarios.

Llama 3 70B Instruct Turbo offers exceptional language understanding and generation capabilities, suitable for the most demanding computational tasks.

Llama 3 8B Instruct Lite is designed for resource-constrained environments, providing excellent balanced performance.

Llama 3 70B Instruct Lite is suitable for environments requiring high performance and low latency.

LLaMA-3 Chat (8B) provides multilingual support, covering a rich array of domain knowledge.

LLaMA-3 Chat (70B) is a powerful chat model that supports complex conversational needs.

LLaMA-2 Chat (13B) offers excellent language processing capabilities and outstanding interactive experiences.

LLaMA-2 provides excellent language processing capabilities and outstanding interactive experiences.

Code Llama is an LLM focused on code generation and discussion, with extensive support for various programming languages, suitable for developer environments.

Gemma 2 continues the design philosophy of being lightweight and efficient.

Gemma Instruct (2B) provides basic instruction processing capabilities, suitable for lightweight applications.

Mistral 7B is a model fine-tuned on demand, providing optimized answers for tasks.

Mistral (7B) Instruct is known for its high performance, suitable for various language tasks.

Mistral 7B is a compact yet high-performance model, adept at handling batch processing and simple tasks like classification and text generation, featuring good reasoning capabilities.

Mixtral 8x7B is a pre-trained sparse mixture of experts model for general text tasks.

Mixtral 8x7B is a sparse expert model that utilizes multiple parameters to enhance reasoning speed, suitable for multilingual and code generation tasks.

Mixtral-8x22B Instruct (141B) is a super large language model that supports extremely high processing demands.

WizardLM 2 is a language model provided by Microsoft AI, excelling in complex dialogues, multilingual capabilities, reasoning, and intelligent assistant tasks.

DeepSeek 67B is an advanced model trained for highly complex conversations.

Qwen2.5 is a new large language model series designed to optimize instruction-based task processing.

Qwen2.5 is a new large language model series designed to optimize instruction-based task processing.

Qwen2 is an advanced general-purpose language model that supports various types of instructions.

DBRX Instruct provides highly reliable instruction processing capabilities, supporting applications across multiple industries.

Upstage SOLAR Instruct v1 (11B) is suitable for refined instruction tasks, offering excellent language processing capabilities.

MythoMax-L2 (13B) is an innovative model suitable for multi-domain applications and complex tasks.

StripedHyena Nous (7B) provides enhanced computational capabilities through efficient strategies and model architecture.

accounts/fireworks/models/llama-v3p3-70b-instruct.description

Yi-Large model, featuring exceptional multilingual processing capabilities, suitable for various language generation and understanding tasks.

Meta Llama 3.3 is a multilingual large language model (LLM) with 70 billion parameters (text input/text output), featuring pre-training and instruction-tuning. The instruction-tuned pure text model of Llama 3.3 is optimized for multilingual conversational use cases and outperforms many available open-source and closed chat models on common industry benchmarks.

Llama 3.2 is designed to handle tasks that combine visual and textual data. It excels in tasks such as image description and visual question answering, bridging the gap between language generation and visual reasoning.

Llama 3.2 is designed to handle tasks that combine visual and textual data. It excels in tasks such as image description and visual question answering, bridging the gap between language generation and visual reasoning.

Llama 3.1 8B is a high-performance model that offers rapid text generation capabilities, making it ideal for applications requiring large-scale efficiency and cost-effectiveness.

Llama 3.1 70B provides enhanced AI reasoning capabilities, suitable for complex applications, supporting extensive computational processing while ensuring efficiency and accuracy.

Llama 3 Groq 8B Tool Use is a model optimized for efficient tool usage, supporting fast parallel computation.

Llama 3 Groq 70B Tool Use offers powerful tool invocation capabilities, supporting efficient processing of complex tasks.

Meta Llama 3 8B delivers high-quality reasoning performance, suitable for diverse application needs.

Meta Llama 3 70B provides unparalleled complexity handling capabilities, tailored for high-demand projects.

Gemma 2 9B is a model optimized for specific tasks and tool integration.

Gemma 7B is suitable for medium to small-scale task processing, offering cost-effectiveness.

Mixtral 8x7B provides high fault-tolerant parallel computing capabilities, suitable for complex tasks.

LLaVA 1.5 7B offers integrated visual processing capabilities, generating complex outputs from visual information inputs.

sonar-reasoning-pro.description

sonar-reasoning.description

sonar-pro.description

sonar.description

Pixtral Large is an open-source multimodal model with 124 billion parameters, built on Mistral Large 2. This is the second model in our multimodal family, showcasing cutting-edge image understanding capabilities.

Ministral 3B is Mistral's top-tier edge model.

Ministral 8B is Mistral's cost-effective edge model.

Mistral 7B is a compact yet high-performance model, excelling in batch processing and simple tasks such as classification and text generation, with good reasoning capabilities.

Mixtral 8x7B is a sparse expert model that leverages multiple parameters to enhance reasoning speed, suitable for handling multilingual and code generation tasks.

Mixtral 8x22B is a larger expert model focused on complex tasks, providing excellent reasoning capabilities and higher throughput.

Codestral Mamba is a language model focused on code generation, providing strong support for advanced coding and reasoning tasks.

jamba-1.5-mini.description

jamba-1.5-large.description

Solar Mini is a compact LLM that outperforms GPT-3.5, featuring strong multilingual capabilities, supporting English and Korean, and providing an efficient and compact solution.

Solar Mini (Ja) extends the capabilities of Solar Mini, focusing on Japanese while maintaining efficiency and excellent performance in English and Korean usage.

Solar Pro is a highly intelligent LLM launched by Upstage, focusing on single-GPU instruction-following capabilities, with an IFEval score above 80. Currently supports English, with a formal version planned for release in November 2024, which will expand language support and context length.

grok-3-mini-beta.description

grok-3-latest.description

grok-2-vision-latest.description

Tongyi Qianwen is a large-scale language model that supports input in various languages, including Chinese and English.

Tongyi Qianwen Plus is an enhanced version of the large-scale language model, supporting input in various languages, including Chinese and English.

Tongyi Qianwen Max is a large-scale language model with hundreds of billions of parameters, supporting input in various languages, including Chinese and English. It is the API model behind the current Tongyi Qianwen 2.5 product version.

Qwen is a large-scale language model that supports long text contexts and dialogue capabilities based on long documents and multiple documents.

qwen-vl-plus.description

Tongyi Qianwen's ultra-large-scale visual language model. Compared to the enhanced version, it further improves visual reasoning and instruction-following abilities, providing a higher level of visual perception and cognition.

qwen-vl-ocr-latest.description

The Tongyi Qianwen Math model is specifically designed for solving mathematical problems.

The Tongyi Qianwen Math model is specifically designed for solving mathematical problems.

The Tongyi Qianwen Coder model.

Tongyi Qianwen code model.

The QwQ model is an experimental research model developed by the Qwen team, focusing on enhancing AI reasoning capabilities.

qvq-72b-preview.description

qwen2.5-7b-instruct.description

qwen2.5-14b-instruct.description

qwen2.5-32b-instruct.description

qwen2.5-72b-instruct.description

qwen2.5-14b-instruct-1m.description

qwen2.5-math-7b-instruct.description

qwen2.5-math-72b-instruct.description

qwen2.5-coder-7b-instruct.description