Tencent Hunyuan Hy4 is the official 1M-context flagship for reasoning, coding and agent workloads, with image understanding.

The flagship K3.1 model from Moonshot AI, a 3-trillion-parameter system with a one-million-token context, native vision and always-on reasoning for coding, agentic and knowledge work.

Kimi K3.1 Pro Highspeed carries the same capabilities as Kimi K3.1 Pro while generating output about two to three times faster, at higher per-token prices.

MiniMax M3.1 Flash is a multimodal reasoning model for coding and agent tasks, with a 1M-token context and adjustable reasoning effort.

DeepSeek's open-weight V4.1 Flash tier: fast, low-cost reasoning and tool use with a 1M-token context.

Qwen Image 2.1 Pro generates images from text and edits them with reference images, billed at a flat price per generated image.

StepFun flagship preview for coding and professional workflows, with a one-million-token context and text, image and video understanding.

Tencent Hunyuan Image 3.5 Preview generates images from text and continues editing them from reference images over multiple turns, billed by output image tokens.

Xiaomi's MiMo-V2.6 Flash for assistant, coding and tool-calling workloads, with a 1M-token context and text, image, video and audio input.

Xiaomi's MiMo-V2.6 Pro for autonomous software engineering, with a 1M-token context and text, image, video and audio input.

Alibaba's Qwen3.8 Omni Flash adds audio to the Qwen3.8 Flash input set, keeping the 1M-token context and text output.

Zhipu AI's GLM-5.3 FlashX: a further GLM-5.3 Flash variant for coding, reasoning and agent workflows, priced between GLM-5.3 Flash and GLM-5.3.

ByteDance's Seed 2.1 Lite model for text, image, video and audio understanding. Supports reasoning, coding and agent workflows, tool calling and structured output.

ByteDance's Seed 2.1 Pro model for complex coding and agent tasks, with enhanced long-running execution, research and multimodal understanding. Supports reasoning, tool calling and structured output.

ByteDance's lowest-priced Seedream 5.0 tier for text-to-image generation and reference-image editing, billed at a flat price per image.

A continuously updated ByteDance model for coding and agent workflows. It supports reasoning, tool calling, structured output, and text, image and video understanding.

MiniMax's H3 Max video model for text- and image-to-video with multi-image, video and audio references and synchronized audio, at 480p or 768p for five to fifteen seconds.

Tencent Hunyuan's open-weight 770B flagship in preview, for reasoning, coding and agent workloads.

Alibaba's lower-cost Qwen3.8 tier for reasoning, coding and agent workflows, with a 1M-token context and text, image and video input.

Zhipu AI's GLM-5.3 Flash: the faster, lower-cost GLM-5.3 tier for coding, reasoning and agent workflows, with a 1M-token context.

Alibaba's Wan 3.0 generates video from text, image, video and audio references at 480p to 1080p for two to thirty seconds.

Wan 3.0 Prime is Alibaba's accelerated video tier, generating from text, image, video and audio references at 480p to 1080p for two to thirty seconds.

Zhipu AI's open-weight GLM-5.3 model for coding, reasoning and agent workflows.

DeepSeek's open-weight V4 flagship for deep reasoning, coding and long-horizon agent tasks.

Qwen image generation and editing with up to three image references and native one- and two-kilopixel output billing.

Qwen Image 3.0 Pro generates and edits images with up to three image references and one- or two-kilopixel output billing, at rates above Qwen Image 3.0.

Alibaba's flagship Qwen reasoning model with a 1M-token context and text, image and video input.

MiniMax's H3 video model for text- and image-to-video with multi-image, video and audio references and synchronized audio, at 768p or 2K for four to fifteen seconds.

MiniMax Speech 2.8 HD synthesizes speech synchronously as MP3 from the supported system voices, billed by the character count the provider reports.

MiniMax Speech 2.8 Turbo is the lower-priced synchronous speech tier, synthesizing MP3 from the supported system voices and billed by the reported character count.

Moonshot AI's largest open-source model, built for deep reasoning and long-context agent workflows.

Tencent Hunyuan's Hy3 for reasoning, coding and agent workloads, with a 256K-token context, tool calling and prompt caching.

A ByteDance Seed 2.1 model balancing capability and cost for production coding and agent workloads. Supports text, image and video understanding, reasoning, tool calling and structured output.

ByteDance's Seedream 5.0 Pro for text-to-image generation and reference-image editing, billed by output size tier at one to two kilopixels.

Alibaba's HappyHorse 1.1 image-to-video model animates a first-frame image for three to fifteen seconds at 480p to 1080p, keeping the input aspect ratio.

Alibaba's HappyHorse 1.1 reference-to-video model generates three- to fifteen-second clips at 480p to 1080p from up to nine reference images.

Alibaba's HappyHorse 1.1 text-to-video model generates three- to fifteen-second clips at 480p, 720p or 1080p, billed by successfully generated output seconds.

Moonshot AI's Kimi K2.7 Code for coding and agent work, with a 256K-token context and text, image and video input.

MiniMax's open-weight coding and agent model with a 1M-token context and native tool use.

StepFun's 198B MoE vision-language model tuned for coding agents and search workflows.

Baidu's ERNIE 5.1 for reasoning and agent workloads, with a 128K-token context, tool calling and structured output.

Tencent Hunyuan Image 3.0 generates and edits images from reference inputs in a single synchronous response, billed by output image tokens.

Kuaishou's Kling V3 for text-to-video and image-to-video with synchronized audio, generating three- to fifteen-second clips from 720p up to 4K.

ByteDance's Seedream 5.0 Lite for text-to-image generation and reference-image editing at two to four kilopixels, billed at a flat price per image.