DeepSeek V3.2 is a MoE model (671B total, 37B active) with strong reasoning, coding, and tool use.

FLUX.2 is an image model (9B) with strong prompt understanding, fast inference, and high-quality photorealistic generation.

GPT-OSS-120B is an open-weight MoE model (117B total, 5.1B active) optimized for single-GPU use, with strong reasoning, tool use, and structured outputs.

GPT-OSS-20B is an open-weight MoE model (21B total, 3.6B active) built for low-latency deployment with strong reasoning, tool use, and structured outputs.

Hunyuan Image 3.0 is a multimodal MoE model (80B) for image generation with strong prompt adherence, world knowledge, and cinematic-quality outputs.

Kimi K2.5 is a multimodal MoE model (1T total, 32B active) with strong visual coding, agentic tool use, and multi-agent orchestration.

MiniMax M2.5 is a MoE model (230B total, 10B active) with strong coding, agentic tool use, and productivity performance, delivering fast, cost-efficient results.

Qwen3.5 is a multimodal MoE model (397B total, 17B active) with a 1M-token context and strong reasoning, coding, and vision-language performance.

Qwen3-ASR is a 1.7B speech model for 52 languages, supporting streaming and offline use with strong noise handling and accurate transcription with timestamps.

Qwen3 Coder is a MoE model (480B total, 35B active) built for agentic coding, with long context, strong tool use, and reliable multi-step code generation.

Qwen3 Instruct is a MoE model (235B total, 22B active) optimized for instruction following, multilingual use, and business apps with strong structured outputs.

Qwen3 VL Instruct is a vision-language MoE model (30B total, 3B active) for visual understanding, OCR, and multimodal agent tasks with strong spatial reasoning.

Qwen-Image is an open-weight diffusion model with strong text rendering, multilingual typography, and image generation and editing.

Qwen-Image-Edit is a diffusion model for image generation and editing, supporting style transfer, object edits, and layout changes with strong prompt control.

Wan2.2-I2V is an open-source video model for image-to-video generation with strong motion consistency and high-quality, realistic outputs.

Wan2.2-T2V is an open-source video model for text-to-video generation with strong motion, coherence, and high-quality outputs.