deepinfra AI Models

Bria/erase — served on DeepInfra's GPU cloud for scalable, cost-efficient inference.

Meta Llama 3.1 70B Instruct on DeepInfra — powerful open-source model for complex reasoning tasks.

Qwen3-Next-80B-A3B-Instruct is an instruction-tuned chat model in the Qwen3-Next series optimized for fast, stable responses without “thinking” traces. It targets complex tasks across reasoning, code generation, knowledge QA, and multilingual...

Qwen3-32B is a dense 32.8B parameter causal language model from the Qwen3 series, optimized for both complex reasoning and efficient dialogue. It supports seamless switching between a "thinking" mode for...

Llama Guard 4 is a Llama 4 Scout-derived multimodal pretrained model, fine-tuned for content safety classification. Similar to previous versions, it can be used to classify content in both LLM...

NVIDIA's Llama 3.1 Nemotron 70B Instruct — fine-tuned for helpfulness and aligned with human preferences.

Qwen3, the latest generation in the Qwen large language model series, features both dense and mixture-of-experts (MoE) architectures to excel in reasoning, multilingual support, and advanced agent tasks. Its unique...

Qwen3-Max is an updated release built on the Qwen3 series, offering major improvements in reasoning, instruction following, multilingual support, and long-tail knowledge coverage compared to the January 2025 version. It...

Gemma 3 introduces multimodality, supporting vision-language input and text outputs. It handles context windows up to 128k tokens, understands over 140 languages, and offers improved math, reasoning, and chat capabilities,...

Gemma 3 introduces multimodality, supporting vision-language input and text outputs. It handles context windows up to 128k tokens, understands over 140 languages, and offers improved math, reasoning, and chat capabilities,...

Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token during inference — delivering near-31B quality at...

This lightweight sentence embedding model excels at converting text into dense vector representations for semantic similarity tasks, making it ideal for developers and researchers needing efficient retrieval, clustering, and matching capabilities with minimal computational overhead.

This advanced language model excels at complex reasoning, long-form content generation, and nuanced analysis tasks, making it ideal for professionals, researchers, and developers requiring high-quality outputs for demanding applications.

This optical character recognition model excels at extracting and digitizing text from documents and images, making it ideal for researchers, developers, and organizations needing to convert scanned papers, PDFs, and photographs into machine-readable text.

This is a high-performance 70-billion parameter language model optimized for complex reasoning tasks and long-form content generation, ideal for developers and enterprises requiring advanced inference capabilities without the computational overhead of larger models.

This advanced multimodal vision-language model excels at understanding and analyzing images alongside text, making it ideal for developers and enterprises needing sophisticated visual reasoning, document processing, and image-based question-answering capabilities.

This fine-tuned 70-billion parameter language model excels at creative writing, roleplay, and narrative generation tasks, making it ideal for authors, game developers, and creative professionals seeking high-quality, imaginative text outputs with minimal safety restrictions.

This advanced image editing model excels at precise object removal, inpainting, and content-aware modifications, making it ideal for designers, photographers, and content creators who need professional-quality image manipulation without manual editing.

A compact multimodal language model optimized for fast inference and lower computational costs, ideal for developers and organizations seeking efficient text and image understanding capabilities without requiring enterprise-grade hardware.

This instruction-tuned 8 billion parameter language model excels at following user directions for tasks like question answering, summarization, and creative writing, making it ideal for developers and organizations seeking a fast, cost-effective alternative to larger models without sacrificing quality on most general-purpose NLP applications.

This lightweight sentence embedding model excels at converting text into dense vector representations for semantic similarity tasks, making it ideal for developers and researchers needing fast, efficient paraphrase detection, clustering, and semantic search with minimal computational overhead.

This multimodal AI model excels at understanding and reasoning about images alongside text, making it ideal for developers and enterprises building applications requiring visual question answering, document analysis, and image-to-text tasks at scale.

This lightweight 0.8 billion parameter language model excels at fast, efficient text generation and reasoning tasks for developers and organizations seeking low-latency inference with minimal computational overhead.

This image editing model is ideal for developers and creative professionals who need to programmatically perform precise object removal, inpainting, and content-aware editing tasks through an API.

A fast and efficient language model optimized for rapid inference and real-time applications, ideal for developers and teams needing quick responses without sacrificing quality across general-purpose NLP tasks.

A lightweight 4-billion parameter language model optimized for efficient inference on resource-constrained devices, ideal for developers and organizations seeking fast, cost-effective AI capabilities without sacrificing reasonable quality across general-purpose tasks.

I don't have verified information about a model called "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8" in my training data, so I cannot provide a factually accurate description of this specific model without risking misinformation.

This image editing model is designed for users who need to perform inpainting, outpainting, and object removal tasks by specifying regions to edit while preserving the rest of the image content.

This lightweight 0.9B parameter vision-language model excels at optical character recognition and document understanding tasks, making it ideal for developers and businesses seeking efficient on-device or edge deployment of text extraction and layout analysis applications.

This large-scale code generation and software development model excels at complex programming tasks, algorithm design, and technical problem-solving, making it ideal for professional developers, software engineers, and technical teams seeking advanced AI-assisted coding capabilities.

This lightweight 2B parameter language model is ideal for developers and organizations seeking fast, efficient inference for real-time applications, chatbots, and edge deployment where computational resources are limited but reasonable language understanding is still required.

This instruction-tuned 32-billion parameter language model excels at following complex directives and generating coherent text across diverse tasks, making it ideal for developers and researchers seeking a capable open-source alternative to proprietary models for applications like content generation, code assistance, and question answering.

This multimodal vision-language model excels at understanding and generating content from both images and text, making it ideal for developers and researchers building applications requiring visual reasoning, image captioning, visual question answering, and text-to-image generation tasks.

Mistral Small 3 is a 24B-parameter language model optimized for low-latency performance across common AI tasks. Released under the Apache 2.0 license, it features both pre-trained and instruction-tuned versions designed...

Alibaba Qwen2.5 72B Instruct on DeepInfra — top open-source model with multilingual and coding strengths.

This lightweight 600M parameter embedding model excels at converting text into vector representations for semantic search, clustering, and retrieval tasks, making it ideal for developers and organizations seeking fast, efficient embeddings with minimal computational overhead.

This lightweight 12 billion parameter vision-language model excels at multimodal tasks combining image and text understanding, making it ideal for developers and organizations seeking efficient inference on resource-constrained hardware without sacrificing capability.

This advanced text-to-image generation model excels at creating high-quality, photorealistic images from detailed prompts and is ideal for designers, content creators, marketers, and developers who need production-ready visual assets quickly.

This powerful large language model excels at complex reasoning, nuanced writing, and code generation tasks, making it ideal for professionals, researchers, and developers requiring high-quality outputs for demanding applications.

This model is optimized for clear, concise text generation and summarization tasks, making it ideal for professionals, content creators, and developers who need reliable outputs for documentation, communication, and knowledge work applications.

This lightweight 9-billion parameter language model is optimized for fast inference and low resource consumption, making it ideal for developers and organizations seeking to deploy efficient AI assistants on edge devices or cost-constrained infrastructure.

This high-performance reasoning model excels at complex problem-solving, mathematical calculations, coding tasks, and logical analysis, making it ideal for developers, researchers, and professionals who need accurate, detailed solutions to challenging computational problems.

This image enhancement model is designed for photographers, content creators, and designers who need to upscale, denoise, and improve the quality of photos through AI-powered restoration techniques.

Mixtral 8×7B Instruct on DeepInfra — popular MoE model with 32K context and strong multilingual performance.

This advanced text-to-image generation model excels at creating photorealistic and highly detailed images from text descriptions, making it ideal for designers, artists, marketers, and content creators who need production-quality visual assets quickly.

Wan-AI/Wan2.7-Image-Edit — served on DeepInfra's GPU cloud for scalable, cost-efficient inference.

Bria/gen_fill — served on DeepInfra's GPU cloud for scalable, cost-efficient inference.

DeepSeek V3 — 671B MoE model with exceptional coding and math performance at very low cost.

This large-scale text embedding model excels at converting documents and queries into dense vector representations for semantic search, retrieval-augmented generation, and similarity matching tasks, making it ideal for developers and researchers building information retrieval systems and AI applications that require precise semantic understanding.

gpt-oss-120b is an open-weight, 117B-parameter Mixture-of-Experts (MoE) language model from OpenAI designed for high-reasoning, agentic, and general-purpose production use cases. It activates 5.1B parameters per forward pass and is optimized...

Qwen3-VL-30B-A3B-Instruct is a multimodal model that unifies strong text generation with visual understanding for images and videos. Its Instruct variant optimizes instruction-following for general multimodal tasks. It excels in perception...

Microsoft Phi-4 14B — small language model achieving state-of-the-art results on reasoning tasks.

Qwen3-14B is a dense 14.8B parameter causal language model from the Qwen3 series, designed for both complex reasoning and efficient dialogue. It supports seamless switching between a "thinking" mode for...

Qwen3-235B-A22B-Thinking-2507 is a high-performance, open-weight Mixture-of-Experts (MoE) language model optimized for complex reasoning tasks. It activates 22B of its 235B parameters per forward pass and natively supports up to 262,144...

Meta Llama 3.1 8B Instruct on DeepInfra — fast, affordable open-source model with 128K context.

Gemma 3 introduces multimodality, supporting vision-language input and text outputs. It handles context windows up to 128k tokens, understands over 140 languages, and offers improved math, reasoning, and chat capabilities,...

Qwen3-Max-Thinking is the flagship reasoning model in the Qwen3 series, designed for high-stakes cognitive tasks that require deep, multi-step reasoning. By significantly scaling model capacity and reinforcement learning compute, it...

One of the highest performing and most popular fine-tunes of Llama 2 13B, with rich descriptions and roleplay. #merge

Qwen3-VL-235B-A22B Instruct is an open-weight multimodal model that unifies strong text generation with visual understanding across images and video. The Instruct model targets general vision-language use (VQA, document parsing, chart/table...

gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for...

Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output. Features a 256K token context window, configurable thinking/reasoning mode, native function...

NVIDIA Nemotron 3 Nano 30B A3B is a small language MoE model with highest compute efficiency and accuracy for developers to build specialized agentic AI systems. The model is fully...