Fireworks' latest Firefunction-v2 is a high-performance function-calling model developed based on Llama-3, optimized for function calls, dialogues, and instruction following.

Fireworks open-source function-calling model provides excellent instruction execution capabilities and customizable features.

fireworks-ai/FireLLaVA-13b is a visual language model that can accept both image and text inputs, trained on high-quality data, suitable for multimodal tasks.

The Llama 3.2 1B instruction model is a lightweight multilingual model introduced by Meta. This model aims to enhance efficiency, providing significant improvements in latency and cost compared to larger models. Sample use cases include retrieval and summarization.

The Llama 3.2 3B instruction model is a lightweight multilingual model introduced by Meta. This model aims to enhance efficiency, providing significant improvements in latency and cost compared to larger models. Sample use cases include querying, prompt rewriting, and writing assistance.

Meta's 11B parameter instruction-tuned image reasoning model. This model is optimized for visual recognition, image reasoning, image description, and answering general questions about images. It understands visual data like charts and graphs, generating text descriptions of image details to bridge the gap between vision and language.

Meta's 90B parameter instruction-tuned image reasoning model. This model is optimized for visual recognition, image reasoning, image description, and answering general questions about images. It understands visual data like charts and graphs, generating text descriptions of image details to bridge the gap between vision and language.

Llama 3.1 8B instruction model, optimized for multilingual dialogues, capable of surpassing most open-source and closed-source models on common industry benchmarks.

Llama 3.1 70B instruction model provides exceptional natural language understanding and generation capabilities, making it an ideal choice for dialogue and analysis tasks.

Llama 3.1 405B instruction model, equipped with massive parameters, suitable for complex tasks and instruction following in high-load scenarios.

Llama 3 8B instruction model, optimized for dialogues and multilingual tasks, delivering outstanding and efficient performance.

Llama 3 70B instruction model, optimized for multilingual dialogues and natural language understanding, outperforming most competitive models.

Llama 3 8B instruction model (HF version), consistent with official implementation results, featuring high consistency and cross-platform compatibility.

Llama 3 70B instruction model (HF version), aligned with official implementation results, suitable for high-quality instruction following tasks.

Mixtral MoE 8x7B instruction model, with a multi-expert architecture providing efficient instruction following and execution.

Mixtral MoE 8x22B instruction model, featuring large-scale parameters and a multi-expert architecture, fully supporting efficient processing of complex tasks.

Mixtral MoE 8x7B instruction model (HF version), performance consistent with official implementation, suitable for various efficient task scenarios.

Phi 3 Vision instruction model, a lightweight multimodal model capable of handling complex visual and textual information, with strong reasoning abilities.

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

Qwen2.5 is a series of decoder-only language models developed by the Alibaba Cloud Qwen team. These models come in different sizes including 0.5B, 1.5B, 3B, 7B, 14B, 32B, and 72B, available in both base and instruct variants.

Qwen2.5 Coder 32B Instruct is the latest version in Alibaba Cloud's series of code-specific large language models. This model significantly enhances code generation, reasoning, and repair capabilities based on Qwen2.5, trained on 55 trillion tokens. It not only improves coding abilities but also maintains advantages in mathematics and general capabilities, providing a more comprehensive foundation for practical applications such as code agents.

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

StarCoder 7B model, trained on over 80 programming languages, boasts excellent code completion capabilities and contextual understanding.

StarCoder 15.5B model supports advanced programming tasks, enhanced multilingual capabilities, suitable for complex code generation and understanding.