Providers that keep no copy of the prompt or the response once the request is finished.
DeepSeek V4.1 Flash Uncensored is an abliterated, uncensored variant of DeepSeek V4.1 Flash, a multimodal Mixture-of-Experts model with 552B backbone parameters (8B active on input, 16B on output) and a 1M-token context window. It natively processes images and text, with strong reasoning, coding, and agentic performance and no content-filtering guardrails.
Kimi K3 is Moonshot AI's open-weight Mixture-of-Experts model with 2.8T total and 104B active parameters and a 1M-token context window. It accepts text, image and video input and produces text, always reasons with low, high or max effort, and is built for long-horizon coding, knowledge work and reasoning.
Mistral Uncensored 1.2 is an uncensored 24B fine-tune of Mistral Small 24B, built on Dolphin Mistral 24B, made for open-ended dialogue, roleplay and creative writing with minimal content restrictions. It accepts text and image input, supports function calling and structured output, and has a 128K-token context window.
DeepSeek V4 Pro is a 1.6T-parameter Mixture-of-Experts model with 49B active parameters and a 1M-token context window. Built for advanced reasoning, coding, and long-horizon agentic workflows with a hybrid attention system for efficient long-context processing.
GLM 5.3 Uncensored is an uncensored variant of Z.ai's GLM-5.3, with refusal behavior removed through abliteration and further fine-tuning. It takes text input and output with a 1M-token context window, always reasons at low, high or max depth, and is aimed at red teaming, security research and agent testing.
Gemma 4 Uncensored is an uncensored variant of Google Gemma 4 26B, a Mixture-of-Experts model with 26B total parameters and only 4B active per token. Fine-tuned for uncensored chat without content filtering, it supports 256K context, coding, and general-purpose conversation.
MiniMax-M3 is MiniMax's natively multimodal Mixture-of-Experts model with about 428B total and 23B active parameters. It takes text, image and video input and produces text, with a context window of up to 1M tokens, and can also operate a desktop computer.
GLM-5.3 is a large-scale reasoning model from Z.ai, built for complex software engineering and long-horizon agent tasks. It supports text input and output with a 1M-token context window, and improves on GLM-5.2 in coding and in the balance between performance and token efficiency.
GLM-4.7-Flash-Heretic is an uncensored experimental variant of GLM-4.7-Flash, optimized for creative freedom and unfiltered dialogue with fast inference speed.
GLM-5.3 is a large-scale reasoning model from Z.ai, built for complex software engineering and long-horizon agent tasks. It supports text input and output with a 1M-token context window, and improves on GLM-5.2 in coding and in the balance between performance and token efficiency.
DeepSeek V4 Flash 0731 is the July 31, 2026 official release of DeepSeek V4 Flash, a Mixture-of-Experts model with 284B total and 13B active parameters and a 1M-token context window. It keeps the preview's architecture and size but is re-post-trained for much stronger agentic performance, with low, high and max reasoning effort.
DeepSeek V4.1 Flash is a multimodal Mixture-of-Experts model with 552B backbone parameters (8B active on input, 16B on output) and a 1M-token context window. It natively processes images and text, with strong reasoning, coding, and agentic performance.
GLM-5.2 is Z.ai's open-weight Mixture-of-Experts model for long-horizon tasks and software engineering. It takes text input and output with a 1M-token context window and up to 128K output tokens, and offers multiple thinking effort levels to balance performance and latency.
Qwen 3.8 27B is an open-weight dense model from Alibaba's Qwen team with 27B parameters and a vision encoder, using hybrid Gated DeltaNet and gated attention layers. It takes text, image and video input with a native 262K-token context window and is built for coding, professional and research work, and long-horizon agentic tasks, with adjustable thinking effort.
Kimi K3 is Moonshot AI's open-weight Mixture-of-Experts model with 2.8T total and 104B active parameters and a 1M-token context window. It accepts text, image and video input and produces text, always reasons with low, high or max effort, and is built for long-horizon coding, knowledge work and reasoning.
Gemma 4 31B is the dense instruction-tuned model in Google DeepMind's open-weight Gemma 4 family, with about 30.7B parameters. It takes text and image input and produces text output, with a 256K-token context window, configurable thinking, native function calling and pretraining on over 140 languages.
Moonshot AI's open-weight multimodal agentic model has 1T total and 32B active parameters, built for long-horizon coding and tool use.
Qwen3.5-397B-A17B is a multimodal Mixture-of-Experts model from Alibaba's Qwen team, with 397B total parameters and 17B active per token. It takes text, image and video input and produces text, with a 262K-token native context window (extendable to about 1M), a default thinking mode, tool calling and support for 201 languages and dialects.
Qwen's dense 27B vision-language model supports reasoning, coding and tool-driven tasks.
GLM-5.1 is Z.ai's open-weight Mixture-of-Experts model for agentic engineering, with 744B total and 40B active parameters. It takes text input and output with a 200K-token context window and up to 128K output tokens, and is tuned for coding agents and long-running tool use.
DeepSeek-V3.2 uses DeepSeek Sparse Attention for efficient long contexts, with strong reasoning and tool use.
Llama 3.3 70B is Meta's multilingual instruction-tuned text model, released in December 2024. It takes text input and produces text and code with a 128K-token context window, officially supports eight languages, and is built for assistant-style chat.
gpt-oss-120b is OpenAI's open-weight Mixture-of-Experts language model with 117B total and 5.1B active parameters, released under the Apache 2.0 license. It takes text input and produces text output with a 128K-token context window, configurable low, medium and high reasoning, full chain-of-thought access and native tool use including function calling and structured outputs.
Voxtral Small 24B is a text model.
Grok 4.20 is xAI's reasoning model built for speed and agentic tool calling, with low hallucination rates and strict prompt adherence. It takes text and image input and produces text output, with a 1M-token context window, function calling and structured outputs.
Grok 4.20 Multi-Agent is a variant of xAI Grok 4.20 designed for collaborative, agent-based workflows. Multiple agents operate in parallel to conduct deep research, coordinate tool use, and synthesize information across complex tasks.
GLM-5.3-Flash is a native multimodal model from Z.ai. It is suited for efficient coding and long-horizon agent tasks. Its hybrid sparse and linear attention architecture maintains accurate long-context behavior while reducing compute overhead.
Kolibri 1 is a text model.
DeepSeek V4 Flash (Preview) is the April 2026 preview release of DeepSeek's efficiency-focused Mixture-of-Experts model, with 284B total and 13B active parameters and a 1M-token context window. It offers a non-thinking mode and two thinking levels (high and max) for fast reasoning and coding workloads.
DeepSeek-V4-Flash-Vision-Exp is DeepSeek's experimental multimodal model in the V4-Flash family, adding visual understanding to the V4-Flash architecture. It serves a 1M-token (1,048,576) context window and supports image input with visual grounding, tool calling, structured/JSON output, and configurable reasoning effort (low/high/max, or disabled).
DeepSeek V4.1 Flash Fast is a lower-latency serving tier of DeepSeek V4.1 Flash, the same model served for faster output. It is a Mixture-of-Experts model with 552B backbone parameters (8B active on input, 16B on output) and a 1M-token context window, with strong reasoning, coding, and agentic performance.
Muse Spark Open is an open-weight reasoning model from Meta for coding and agentic workflows. It supports tool calling and native image understanding in a single checkpoint, using a sparse Mixture-of-Experts architecture with 816B total / 42B active parameters and a 1M-token context window.
Hy4 Preview is a 770B-parameter Mixture-of-Experts model from the Tencent Hy Team, with 49B parameters active per token and a 1M-token context window. It uses Gated DeepSeek Sparse Attention with IndexCache and is post-trained for software engineering, office and data analysis, game development and scientific research, with adjustable reasoning effort.
Hy4 Preview is a 770B-parameter Mixture-of-Experts flagship model from the Tencent Hy team, with 49B parameters active per token and a 1M-token context window. It is post-trained for software engineering, office and data analysis, game development, and scientific research, with strong long-horizon reasoning, tool use, and agentic performance.
MiMo-V2.6-Pro is the flagship of Xiaomi's MiMo-V2.6 series: a sparse Mixture-of-Experts model with 1.02T total and 42B activated parameters, natively omnimodal across text, image, video and audio, with a 1M-token context window. Trained with large-scale mixed reinforcement learning and groupwise agentic grading, it is built for demanding agentic work including long-horizon coding, multi-tool agents and cybersecurity tasks.
GLM-5.2 Fast is a lower-latency serving tier of GLM-5.2, Z.ai's model for long-horizon tasks, served for faster output. It takes text input and output with a 1M-token context window, and offers stronger coding than GLM-5.1 with multiple thinking effort levels to balance performance and latency.
GLM-5.3 Fast is a lower-latency serving tier of GLM-5.3, Z.ai's reasoning model for complex software engineering and long-horizon agent tasks. It shares GLM-5.2's base model with improved post-training for coding, and supports text input and output with a 1M-token context window.
GLM-5.3-Flash is a native multimodal model from Z.ai. It is suited for efficient coding and long-horizon agent tasks. Its hybrid sparse and linear attention architecture maintains accurate long-context behavior while reducing compute overhead.
GLM 5.3 Flash Uncensored is a text model.
DeepSeek V4.1 Flash is a multimodal Mixture-of-Experts model with 552B backbone parameters (8B active on input, 16B on output) and a 1M-token context window. It natively processes images and text, with strong reasoning, coding, and agentic performance.
DeepSeek V4 Pro is a 1.6T-parameter Mixture-of-Experts model with 49B active parameters and a 1M-token context window. Built for advanced reasoning, coding, and long-horizon agentic workflows with a hybrid attention system for efficient long-context processing.
DeepSeek V4 Flash 0731 Fast is a faster serving variant of DeepSeek V4 Flash 0731, DeepSeek's July 2026 release of its 284B-parameter Mixture-of-Experts model with 13B active parameters. The model has a 1M-token context window and is re-post-trained for stronger agentic performance, with low, high and max reasoning effort.
DeepSeek V4 Pro is a 1.6T-parameter Mixture-of-Experts model with 49B active parameters and a 1M-token context window. Built for advanced reasoning, coding, and long-horizon agentic workflows with a hybrid attention system for efficient long-context processing.
DeepSeek V4.1 Flash is a multimodal Mixture-of-Experts model with 552B backbone parameters (8B active on input, 16B on output) and a 1M-token context window. It natively processes images and text, with strong reasoning, coding, and agentic performance.
DeepSeek V4.1 Flash 1017 Uncensored is an abliterated, uncensored variant of DeepSeek V4.1 Flash 1017, a multimodal Mixture-of-Experts model with 552B backbone parameters (8B active on input, 16B on output) and a 1M-token context window. It natively processes images and text, with strong reasoning, coding, and agentic performance and no content-filtering guardrails.
Kimi K3 is an ultra-large-scale, open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at navigating large repositories, using tools, debugging, and iterating against images, logs, tests, and runtime feedback.
Grok 4.3 is xAI's fast, reliable model with strong tool calling and instruction following, suited to agentic workflows. It takes text and image input and produces text output, with a 1M-token context window, function calling, structured outputs and reasoning effort from none to xhigh.
MiMo-V2.6-Flash is Xiaomi's efficiency-balanced omnimodal model: a sparse Mixture-of-Experts with 309B total and 15B activated parameters, natively understanding text, image, video, and audio with a 1M-token context window. Trained with mixed reinforcement learning across coding, general agents, visual, and cybersecurity tasks, it is built for long-horizon coding, tool use, and computer use.
MiniMax-M3 is MiniMax's natively multimodal Mixture-of-Experts model with about 428B total and 23B active parameters. It takes text, image and video input and produces text, with a context window of up to 1M tokens, and can also operate a desktop computer.
Inkling Small is a smaller model in Thinking Machines Lab's Inkling family, a Mixture-of-Experts with 276B total and 12B active parameters. Like Inkling, it accepts text, image and audio input and generates text, with a context of up to 1M tokens and variable thinking effort.
Grok 4.5 is xAI's intelligent coding model for agentic software engineering and workflow tasks, with function calling, structured outputs, and a 500K-token context window.
Grok 4.6 is xAI's model for coding, agentic tasks and knowledge work, designed for long, multi-step work such as research, codebase tasks and turning an idea into a working application. It takes text and image input and produces text output, with a 500K-token context window, function calling, structured outputs and low to xhigh reasoning effort.
Grok 4.7 is xAI's model for software engineering and professional knowledge work, especially long-running tasks. Compared with Grok 4.6 it uses a larger base model trained longer on harder tasks and checks its own output more carefully, and it takes text and image input with a 500K-token context window and low to xhigh reasoning effort.
Llama 4 Scout is Meta's natively multimodal Mixture-of-Experts model with 17B active parameters, 16 experts and 109B total parameters. It takes text and image input and produces text and code with a context window of up to 10M tokens, and is built for assistant chat and visual reasoning such as captioning and image question answering.
Qwen 3.8 2.4T is Alibaba's open-weight 2.4-trillion-parameter MoE model (95B active), with major gains in software engineering, research, and long-horizon agentic tasks. It is text-only, requires thinking mode, and supports a 262K-token context window.
Gemma 4 31B Instruct Turbo is a faster serving variant of Gemma 4 31B, the dense instruction-tuned model in Google DeepMind's open-weight Gemma 4 family. The base model takes text and image input and produces text output, with a 256K-token context window, configurable thinking and native function calling.
Ling 3.1 Flash is inclusionAI's native hybrid reasoning model, a hybrid-linear Mixture-of-Experts with 560B total and 25B active parameters that builds on Ling 3.0 Flash. It takes text input and output with a 1M-token context window, is designed for complex real-world tasks, and lets thinking be turned on or off.
NVIDIA Nemotron 3.5 Lightning is NVIDIA's fastest open model for always-on agents and high-volume specialized tasks. It delivers a substantial leap in agentic capability over its predecessor Nemotron 3 Nano, with up to 4x higher throughput on a 1M-token context.
NVIDIA Nemotron 3.5 Super VL is an open multimodal model for agents that see, reason, and act. Its 120B-parameter Mixture-of-Experts design (12B active) understands text, images, documents, GUIs, and video, and adds reasoning and tool use on top.
Qwen3.5-35B-A3B is an efficient Mixture-of-Experts model from Alibaba's Qwen3.5 series with 35B total parameters and only 3B activated per token. It features a 262K token context window (extensible to 1M with YaRN), thinking/reasoning mode, tool calling, and support for 201 languages. Delivers strong performance on reasoning, coding, and vision-language tasks at a fraction of the compute cost.
Qwen3.5-9B is a high-performance model from Alibaba's Qwen3.5 series with a hybrid Gated Delta Networks and sparse MoE architecture. It features a 262K token context window, thinking/reasoning mode, tool calling, multi-token prediction, and support for 201 languages. Excels at reasoning, coding, instruction following, and long-context tasks.
Qwen 3.8 2.4T is Alibaba's open-weight 2.4-trillion-parameter MoE model (95B active), with major gains in software engineering, research, and long-horizon agentic tasks. It is text-only, requires thinking mode, and supports a 262K-token context window.
Qwen3 VL 30B A3B Instruct is a Mixture-of-Experts vision-language model from Alibaba's Qwen team. It takes text, image and video input and produces text with a native 256K-token context window, and handles visual agent tasks, visual coding, 2D and 3D grounding, long video understanding and OCR in 32 languages.
Qwen3.5-27B is Alibaba's largest dense Qwen3.5 model, delivering near-frontier quality across reasoning, coding, and instruction following. It features a 262K token context window (extensible to 1M), thinking/reasoning mode, tool calling, multi-token prediction, and support for 201 languages. Best suited for production deployments and complex enterprise tasks requiring top-tier performance.
Gemma 4 31B Instruct NVFP4 is a 4-bit NVFP4 quantization of Gemma 4 31B IT, Google DeepMind's 30.7B-parameter dense open model, with both weights and activations quantized. It supports text and image input, configurable thinking, and function calling.
Qwen 3 Coder 480B Turbo is a faster serving variant of Qwen3 Coder 480B A35B Instruct, Alibaba's Qwen team's Mixture-of-Experts coding model with 480B total and 35B active parameters. The model has a native 262K-token context window and is built for agentic coding, tool calling and repository-scale work.
Gemma 4 26B A4B Instruct is the Mixture-of-Experts model in Google DeepMind's open-weight Gemma 4 family, with 25.2B total and 3.8B active parameters per token. It takes text and image input and produces text output, with a 256K-token context window, configurable thinking and native function calling.
Devstral 2 is Mistral AI's 123B dense transformer for agentic software engineering, released with open weights under a modified MIT license. It takes text input and produces text output, with a 256K context window. It is built to power coding agents that explore codebases, call tools and make coordinated edits across multiple files, for work such as bug fixing and modernizing legacy systems.
Mistral Large 3 is Mistral AI's open-weight multimodal granular Mixture-of-Experts model, with 675B total and 41B active parameters. It takes text and image input with a 256K-token context window and is built for agentic work, tool use, retrieval-augmented systems, long documents and coding across dozens of languages.
Mistral Small 3.2 is a 24B open-weight model from Mistral AI that accepts text and image input, released in June 2025 as a minor update to Mistral Small 3.1. It improves instruction following, produces fewer repetitive or infinite generations, and has a more robust function-calling template.
Kimi K2 Thinking is Moonshot AI's open-weight thinking model, a Mixture-of-Experts model with about 1 trillion total parameters and 32 billion active per token. It takes text input and produces text output, with a 256K-token context window. It is built to interleave step-by-step reasoning with function calls, and is designed to stay goal-directed across 200 to 300 sequential tool calls for research, coding and writing tasks.
Kimi K2.5 is Moonshot AI's open-weight, natively multimodal Mixture-of-Experts model with 1T total and 32B active parameters and a 256K-token context window. It accepts text and image input, offers thinking and instant modes, and is built for coding, visual coding, agent swarm workflows and office tasks.
Kimi K2.7 Code is Moonshot AI's coding-focused agentic model built on Kimi K2.6, with 1T total parameters and 32B active parameters. It always operates in thinking mode, supports text and image input, and targets long-horizon software engineering, agentic task decomposition, and multi-turn coding workflows with 256K context.
NVIDIA Nemotron 3 Super is a hybrid Mixture-of-Experts (MoE) model engineered for highest compute efficiency and accuracy in multi-agent applications and specialized agentic systems. It is optimized to run many collaborating agents per application on a single GPU, delivering high accuracy for reasoning, tool use, and instruction following.
NVIDIA Nemotron 3 Ultra is built for frontier reasoning, orchestration, coding agents, deep research, and complex enterprise workflows. It delivers up to 5x faster inference and up to 30% lower cost for agentic workloads.
Qwen 3 Coder 480B Turbo is a faster serving variant of Qwen3 Coder 480B A35B Instruct, Alibaba's Qwen team's Mixture-of-Experts coding model with 480B total and 35B active parameters. The model has a native 262K-token context window and is built for agentic coding, tool calling and repository-scale work.
Qwen 3 Next 80B is a Mixture-of-Experts model from Alibaba's Qwen team with 80B total and 3B active parameters, using hybrid Gated DeltaNet and gated attention layers with multi-token prediction for efficient long-context work. It takes text input and output with a native 262K-token context window.
Qwen 3.6 35B A3B is a fast mixture-of-experts model with 35B total parameters and ~3B active per token. Strong at agentic coding, STEM reasoning, and tool use, with a native 256K context window.
Qwen3 235B A22B Instruct 2507 is a text model.
Qwen3 Coder 30B A3B Instruct is an open-weight Mixture-of-Experts coding model from the Qwen team. It has 30.5B total parameters, with 3.3B active per token (8 of 128 experts), and a native 262,144-token context window. It is a text-only, non-thinking model built for agentic coding, agentic browser use and repository-scale code understanding, with a dedicated function-calling format, and is released under Apache 2.0.
Qwen3 Coder 480B A35B Instruct is a text model.
Qwen3 Coder Next is an open-weight coding model from Alibaba's Qwen team, built for coding agents and local development on the Qwen3-Next hybrid-attention Mixture-of-Experts architecture with 80B total and 3B active parameters. It has a native 262K-token context window and is aimed at agentic coding, tool calling and recovery from execution failures.
Qwen3 Next 80B A3B Instruct is a text model.
xAI's fast coding model trained specifically for agentic coding, currently in early access.
GLM 4.7 Flash is a text model.
Gemma 3 27B Instruct is the 27-billion-parameter instruction-tuned model in Google DeepMind's Gemma 3 family of open-weight multimodal models. It takes text and image input and produces text output, with a 128K-token context window and support for over 140 languages, for tasks such as chat, summarization, question answering and image understanding.
MiniMax-M2.5 is MiniMax's open-weight large language model for coding, agentic tool use and search, and office work. It takes text input and produces text output, and is built for complex multi-step agentic tasks and deliverable outputs such as Word, PowerPoint and Excel documents.
MiniMax-M2.7 is MiniMax's open-weight model of about 229B parameters, built for agentic and professional work. It handles software engineering (debugging, refactoring, code security), complex agent harnesses with tool search, and editing of Word, Excel and PowerPoint files.
GLM-4.6 is Z.ai's open-weight Mixture-of-Experts model with 355B total and 32B active parameters. It takes text input and output with a 200K-token context window and up to 128K output tokens, and is built for agentic coding, front-end development, reasoning, writing and role-play.
GLM-4.7 is Z.ai's open-weight Mixture-of-Experts model with 355B total and 32B active parameters. It takes text input and output with a 200K-token context window and up to 128K output tokens, and is built for agentic coding, UI generation, tool use, reasoning and writing.
GLM-5 is Z.ai's open-weight Mixture-of-Experts model with 744B total and 40B active parameters, built for complex systems engineering and long-horizon agentic tasks. It takes text input and output with a 200K-token context window and up to 128K output tokens.
MiniMax-M2 is MiniMax's open-weight Mixture-of-Experts language model built for coding and agentic workflows, with 230 billion total parameters and 10 billion active per token. It takes text input and produces text output, with a context window of 204,800 tokens. It is designed for multi-file edits, code-run-fix loops and test-validated repairs, and for long tool-calling sequences across shell, browser, retrieval and code-execution tools.
MiniMax-M2.1 is MiniMax's open-weight Mixture-of-Experts language model, an update to MiniMax-M2 focused on real-world coding and agentic tasks. It takes text input and produces text output. MiniMax highlights multi-language programming across Rust, Java, Go, C++, Kotlin, Objective-C, TypeScript and JavaScript, native Android and iOS development, web and app development, tool use and long-horizon planning, and multi-step office workflows.
The DeepSeek R1 model has undergone a minor version upgrade, with the current version being DeepSeek-R1-0528.
DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2.
DeepSeek V3.1 is DeepSeek's hybrid Mixture-of-Experts model with 671B total and 37B active parameters, supporting both thinking and non-thinking modes in a single model. It has a 128K-token context window and improves tool calling and agent tasks over DeepSeek V3.
Llama Guard 4 is a natively multimodal safety classifier with 12 billion parameters trained jointly on text and multiple images. Llama Guard 4 is a dense architecture pruned from the Llama 4 Scout pre-trained model and fine-tuned for content safety classification. Similar to previous versions, it can be used to classify content in both LLM inputs (prompt classification) and in LLM responses (response classification). It itself acts as an LLM: it generates text in its output that indicates whether a given prompt or response is safe or unsafe, and if unsafe, it also lists the content categories violated.
Gemma 3 12B Instruct is the 12-billion-parameter instruction-tuned model in Google DeepMind's Gemma 3 family of open-weight multimodal models. It takes text and image input and produces text output, with a 128K-token context window and support for over 140 languages, for tasks such as chat, summarization, question answering and image understanding.
Gemma 3 4B Instruct is the 4-billion-parameter instruction-tuned model in Google DeepMind's Gemma 3 family of open-weight multimodal models. It takes text and image input and produces text output, with a 128K-token context window and support for over 140 languages, for tasks such as chat, summarization, question answering and image understanding.
Granite 4.2 8B is IBM's 8B-parameter dense reasoning model in the open-weight Granite 4.2 family, for reasoning, code generation, tool calling and agentic workflows. It takes text input and produces text output with a 128K-token native context, and offers full thinking, non-thinking and low-effort thinking modes per request.
Ling 3.0 Flash is inclusionAI's hybrid reasoning model, a hybrid-linear Mixture-of-Experts with 124B total and 5.1B active parameters built for efficient agentic workflows such as coding and deep research. It takes text input and output with a 256K-token context window, and its thinking mode can be turned off per request.
Ling 3.0 Flash VL is inclusionAI's multimodal version of Ling 3.0 Flash, a Mixture-of-Experts model with 124B total and 5.5B active parameters. It takes text, image and video input and produces text with a context of up to 256K tokens, and is built for document and chart understanding, GUI agents and long-video question answering.
Meta Llama 3.1 8B Instruct Turbo is a faster serving variant of Llama 3.1 8B Instruct, Meta's small multilingual instruction-tuned text model. The model has a 128K-token context window, officially supports eight languages and supports tool use.
Llama 3.3 70B Instruct is Meta's 70B-parameter multilingual instruction-tuned model, tuned for assistant-style dialogue in English, German, French, Italian, Portuguese, Hindi, Spanish and Thai. It takes text input and produces text and code with a 128K-token context window. This Turbo variant is a faster serving variant of Llama 3.3 70B Instruct.
Muse Glimmer 30B is an open-weight multimodal model from Meta Superintelligence Lab, distilled from Muse Spark. It is a dense 30B transformer that takes text and image input and produces text with a context window of at least 131K tokens, built for local and coding agents, tool calling and multimodal reasoning on a single consumer GPU or Mac.
12B model trained jointly by Mistral AI and NVIDIA, it significantly outperforms existing models smaller or similar in size.
Nemotron Content Safety 3.5 is a multimodal safety classifier developed by NVIDIA. A compact safety model that handles text, images, and custom policies. It outputs a safe/unsafe classification plus a reasoning trace, and can be used as an inference-time guardrail, as a judge for LLM safety testing and evaluation, or with the accompanying training dataset to post-train models for safer behavior.
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 lower-latency inference. The model is trained in OpenAI’s Harmony response format and supports reasoning level configuration, fine-tuning, and agentic capabilities including function calling, tool use, and structured outputs.
L3.1 70B Euryale v2.2 is Sao10K's fine-tune of Meta Llama 3.1 70B Instruct for creative writing and roleplay. Compared with v2.1 it was trained on more roleplay and creative writing data, with added data for system-prompt adherence and reasoning.
Llama 3.2 3B is a small multilingual text model from Meta, instruction tuned for assistant-style chat and agentic tasks such as knowledge retrieval, summarization and prompt rewriting. It takes text input and produces text and code with a 128K-token context window, and officially supports eight languages.
Magistral Small 1.2 (2509) is Mistral AI's 24B open-weight reasoning model, fine-tuned from Mistral Small 3.2. It takes text and image input and produces text output, generating a reasoning trace before its final answer, with a 128K context window and support for 24 languages. It is built for multi-step reasoning tasks and is compact enough for local deployment once quantized.
Ministral 3 14B Instruct is the largest instruction-tuned model in Mistral AI's Ministral 3 family of dense open-weight models, combining a 13.5B language model with a 0.4B vision encoder. It takes text and image input and produces text output, with native function calling and JSON output, and multilingual support. It is built for private chat and assistant deployments, local agentic workflows and fine-tuning.
Ministral 3 3B Instruct is the instruction-tuned 3B model in Mistral AI's Ministral 3 family of small dense open-weight models, combining a 3.4B language model with a 0.4B vision encoder. It takes text and image input and produces text output, with native function calling and JSON output, and support for 11 languages. It is built for edge and local deployment, and for tasks such as image captioning, classification, translation and data extraction.
Ministral 3 8B Instruct is the instruction-tuned 8B model in Mistral AI's Ministral 3 family of small dense open-weight models, combining an 8.4B language model with a 0.4B vision encoder. It takes text and image input and produces text output, with native function calling and JSON output, and multilingual support. It is built for local assistants, chat in constrained environments, image and document understanding, and agentic use cases.
Mistral Uncensored Roleplay is an uncensored 24B fine-tune of Mistral Small 24B for character-driven roleplay, immersive storytelling and open-ended creative writing with minimal content restrictions. It accepts text and image input, supports function calling and has a 128K-token context window.
NVIDIA Nemotron 3 Nano 30B is a compact and efficient language model from NVIDIA, optimized for fast inference while maintaining strong performance across diverse tasks.
NVIDIA Nemotron Nano 12B v2 VL is an open 12.6B-parameter vision-language model from NVIDIA. It pairs a CRadioV2-H vision encoder with the Nemotron Nano 12B v2 language model. It takes English text, up to four images or a video as input and returns text, with 128K tokens of combined input and output context. It is built for document intelligence on invoices, receipts and manuals, visual question answering, summarization, multi-image reasoning and video understanding.
NVIDIA Nemotron Nano 9B v2 is an open 9B-parameter reasoning and chat model from NVIDIA, built on a hybrid Mamba-2 and Transformer architecture with only four attention layers. It was produced by pruning and aligning the 12B Nemotron Nano v2 base model. It handles text input and output with up to 128K tokens of context and supports tool calling. Reasoning is on by default and can be switched off, and an optional thinking budget caps reasoning tokens.
gpt-oss-safeguard-120b is an open-weight safety reasoning model from OpenAI, fine-tuned from gpt-oss-120b, with 117B total parameters and 5.1B active. It classifies content against a safety policy that the developer writes and supplies at inference time, and it produces a reviewable chain of thought for each decision. It is built for trust and safety tasks such as input and output filtering and content labeling, supports low, medium and high reasoning effort, and is released under Apache 2.0.
gpt-oss-safeguard-20b is the smaller open-weight safety reasoning model from OpenAI, fine-tuned from gpt-oss-20b, with about 21B total parameters and 3.6B active. It reads a developer-written safety policy at inference time and classifies content against it with a visible chain of thought, for lower-latency trust and safety tasks such as input and output filtering and content labeling. It supports low, medium and high reasoning effort and is released under Apache 2.0.
Qwen 3 235B A22B Instruct 2507 is the July 2025 update of Alibaba's Qwen3 Mixture-of-Experts model, with 235B total and 22B active parameters. This variant runs in non-thinking mode only, answering directly without reasoning blocks, and has a native 262K-token context window.
Qwen 3 235B A22B Thinking 2507 is the July 2025 update of Alibaba's Qwen3 Mixture-of-Experts model, with 235B total and 22B active parameters. This variant always thinks before answering, for complex reasoning tasks, and has a native 262K-token context window.
Qwen3-VL 235B A22B Instruct is Alibaba's Qwen vision-language Mixture-of-Experts model with about 236B total and 22B active parameters. It takes image, video and text input and produces text, with a native 256K-token context, OCR in 32 languages, GUI agent abilities and long-video understanding.
Qwen3 14B is a dense 14.8B-parameter model in the Qwen3 family from Alibaba's Qwen team, released in April 2025. It switches between a thinking mode for complex reasoning and a fast non-thinking mode, supports 119 languages, and has a native 32K-token context window (131K with YaRN).
Qwen3 30B A3B is a Mixture-of-Experts model in the Qwen3 family with 30.5B total and 3.3B active parameters from Alibaba's Qwen team, released in April 2025. It switches between a thinking mode for complex reasoning and a fast non-thinking mode, supports 119 languages, and has a native 32K-token context window (131K with YaRN).
Qwen3 32B is a dense 32.8B-parameter model in the Qwen3 family from Alibaba's Qwen team, released in April 2025. It switches between a thinking mode for complex reasoning and a fast non-thinking mode, supports 119 languages, and has a native 32K-token context window (131K with YaRN).
Gemma 4 12B Instruct is the instruction-tuned 12B model in Google DeepMind's Gemma 4 family of open models. It is a dense 11.95B-parameter multimodal model with configurable thinking modes, native function calling, and support for over 140 languages.
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 for efficient local deployment. The model achieves 81% accuracy on the MMLU benchmark and performs competitively with larger models like Llama 3.3 70B and Qwen 32B, while operating at three times the speed on equivalent hardware.
Llama Nemotron Embed VL 1B v2 is NVIDIA's multimodal embedding model for question-answering retrieval over documents. It pairs a Llama 3.2 1B language model with a SigLIP2 image encoder (about 1.7B parameters total) and embeds text queries and document pages given as images, text or both, including tables, charts and infographics, as 2,048-dimensional vectors.
Qwen2.5 is a model pretrained on a large-scale dataset of up to 18 trillion tokens, offering significant improvements in knowledge, coding, mathematics, and instruction following compared to its predecessor Qwen2. The model also features enhanced capabilities in generating long texts, understanding structured data, and generating structured outputs, while supporting multilingual capabilities for over 29 languages.
Qwen3 Embedding 0.6B is the smallest text embedding model in Alibaba's Qwen team's Qwen3 Embedding series, with 28 layers and a 32K-token context. It supports more than 100 natural and programming languages, takes optional task instructions, and produces embeddings of up to 1,024 dimensions that can be set to a smaller size.
Qwen3 Embedding 8B is the largest text embedding model in Alibaba's Qwen team's Qwen3 Embedding series, with 36 layers and a 32K-token context. It supports more than 100 natural and programming languages, takes optional task instructions, and produces embeddings of up to 4,096 dimensions that can be set anywhere from 32 to 4,096.
Voxtral Mini 3B is a text model.
Phi-4 is Microsoft's 14B-parameter dense decoder-only language model, trained on a blend of synthetic data, filtered public web content and academic books and Q&A data with a focus on reasoning. It takes text input and produces text output with a 16K-token context window and is intended primarily for English.
BGE-EN-ICL is an English text embedding model from BAAI (Beijing Academy of Artificial Intelligence), built on a 7B-parameter Mistral architecture. It supports in-context learning: adding a task instruction and a few example query and answer pairs to a query helps it adapt to new retrieval tasks, and it also works zero-shot.
BGE-M3 is a multilingual text embedding model from BAAI (Beijing Academy of Artificial Intelligence), built from XLM-RoBERTa-large. It supports more than 100 languages and inputs up to 8,192 tokens, and one model provides dense, sparse (lexical) and multi-vector retrieval, producing 1,024-dimensional dense embeddings.
L3 8B Lunaris v1 Turbo is a faster serving variant of Lunaris v1, Sao10K's 8B generalist and roleplay model. Lunaris is a merge of four Llama 3 8B fine-tunes on Meta Llama 3 8B Instruct, aimed at better creativity and logic than Sao10K's Stheno v3.2.
Palmyra Vision 7B is a text model.
Multilingual E5 Large Instruct is a text embedding model from Microsoft researchers, initialized from XLM-RoBERTa-large with 24 layers and 1,024-dimensional embeddings. It covers about 100 languages and takes inputs up to 512 tokens, with each query paired with a one-sentence task instruction to guide retrieval and similarity tasks.
Z-Image Turbo is the distilled, speed-focused variant of Z-Image, a 6B-parameter open-weight image model from Alibaba's Tongyi-MAI team built on a single-stream diffusion transformer. It generates images from a text prompt in 8 sampling steps, with photorealistic output and English and Chinese text rendering.
Chroma1-HD is an 8.9B-parameter open-weight text-to-image model by lodestones (Lodestone Rock), derived from Black Forest Labs' FLUX.1 [schnell] and released under Apache 2.0. It is intended as a neutral base for fine-tuning and generates images from a text prompt, including adult content.
Orpheus TTS is an open-source text-to-speech model from Canopy Labs, built on Llama 3.2 3B as a speech language model. It produces human-like speech with natural intonation, supports emotion control through simple tags and zero-shot voice cloning, and streams audio for real-time use.
Kokoro is an open-weight text-to-speech model from hexgrad with 82 million parameters, built on the StyleTTS 2 and ISTFTNet architectures. Released under Apache 2.0, it offers 54 voices across 8 languages and is designed to give natural speech while staying fast and inexpensive to run.
Qwen3-TTS 0.6B is the smaller text-to-speech model in Alibaba's Qwen team's Qwen3-TTS series, built on a multi-codebook language model with a 12Hz speech tokenizer. It speaks 10 languages including Chinese, English, Japanese and Spanish, supports streaming output, and balances quality with efficiency, with variants for preset voices and 3-second voice cloning.
Qwen3-TTS 1.7B is the larger text-to-speech model in Alibaba's Qwen team's Qwen3-TTS series, built on a multi-codebook language model with a 12Hz speech tokenizer. It speaks 10 languages including Chinese, English, Japanese and Spanish, supports streaming output and instruction-based style control, with variants for preset voices, voice design from a text description and 3-second voice cloning.
Chatterbox HD is a text-to-speech model from Resemble AI's Chatterbox family. The open-source Chatterbox models are built on a 0.5B Llama backbone and support zero-shot voice cloning from a short reference clip and control over emotional intensity, with a PerTh watermark in generated audio.
WAI-Illustrious-SDXL is a community anime checkpoint by the Civitai creator WAI0731, fine-tuned from OnomaAI's Illustrious XL, an SDXL-based model. It generates anime and illustration-style images from a text prompt, including adult content.
Lustify SDXL is a community fine-tune of SDXL by the Civitai creator coyotte for photorealistic images, including adult content. It understands both natural-language prompts and Danbooru-style tags, and generates images from a text prompt.
Lustify v7 (GGWP) is a release of Lustify, a community fine-tune of SDXL by the Civitai creator coyotte for photorealistic images, including adult content. It understands both natural-language prompts and Danbooru-style tags, and generates images from a text prompt.
Lustify v8 (APEX) is a release of Lustify, a community fine-tune of SDXL by the Civitai creator coyotte for photorealistic images, including adult content. It understands both natural-language prompts and Danbooru-style tags, and generates images from a text prompt.
Stable Diffusion 3.5 Large is Stability AI's 8B-parameter Multimodal Diffusion Transformer (MMDiT) text-to-image model, with improved image quality, typography and complex prompt understanding. It generates images from a text prompt.
HunyuanImage 3.0 is Tencent's open-weight native multimodal image model, unifying understanding and generation in one autoregressive framework. It uses a Mixture-of-Experts design with 80B total and 13B active parameters, generates images from a text prompt and can use reasoning to expand sparse prompts.
MiMo-V2.5 is Xiaomi's native omnimodal model with strong agentic capabilities, supporting text, image, video, and audio understanding in a unified architecture. Built on a sparse Mixture-of-Experts backbone with 310B total and 15B active parameters, it delivers long-context reasoning up to 1M tokens, function calling, and multimodal perception.
