Alibaba Cloud

Alibaba: HappyHorse 1.0 is described as an open source, state-of-the-art AI video generator with native audio and video co-generation capabilities - meaning that the alibaba: HappyHorse 1.0 video model can simultaneously generate video frames and corresponding audio tracks (dialogue, ambience, foley) in a single forward pass, rather than generating silent video first and then dubbing it later. According to an architectural description compiled by the community, the model is built around a 15 billion-parameter unified self-attention Transformer that can process text, image, video, and audio tokens within a single token sequence. It is reportedly built without a dedicated cross-attention branch and without a separate audio module. Combined with DMD-2 distillation technology, its distilled version is said to require only 8 denoising steps on the NVIDIA H100 and requires no classifier-free guidance to generate 1080p video in approximately 38 seconds.

Input:

Output:

Input:

—

Output:

$0.112-0.192/second

Context length:

—

Max output:

—

Alibaba: HappyHorse 1.1 is a video generation model from Alibaba. It generates short videos from a text prompt, a single starting image, or a set of reference images, with output up to 1080p and durations of 3 to 15 seconds. It is suited for creative content, social media clips, and image-driven animation, and improves on the prior version with stronger prompt adherence, smoother motion, and more consistent characters across frames.

Input:

Output:

Input:

—

Output:

$0.049-0.126/second

Context length:

—

Max output:

—

Qwen: Qwen3.7-Flash is a visual language reasoning model from Alibaba. It is suitable for multi-modal agents, visual encoding, search and computer interaction, with advantages in object recognition, spatial understanding and real-world visual perception.

Input:

Output:

Input:

$0.03$0.0195/1M

Output:

$0.13$0.0845/1M

Context length:

1M

Max output:

64K

Qwen3.7-Max is the flagship model of Alibaba’s Qwen3.7 series. It supports text input and output and is designed for agent-centric workloads, with significant advantages in coding, office and productivity tasks, and long-term autonomous execution. The model offers significant improvements in encoding and agent performance compared to previous Qwen versions, and supports explicit cue word caching for efficient repeated context use.

Input:

Output:

Input:

$2.5$1.625/1M

Output:

$7.5$4.875/1M

Context length:

1M

Max output:

64K

Qwen3.7-Plus is a multi-modal generative model in the Qwen3.7 series that supports text, image and video input and text output. It is designed for agent workloads and delivers strong performance in programming, office productivity, and long-range autonomous execution.

Input:

Output:

Input:

$0.4$0.26/1M

Output:

$1.6$1.04/1M

Context length:

1M

Max output:

64K

Alibaba: Wan3.0-Video is a multi-functional reference video generation model that supports text-to-video, image-to-video (first frame/first and last frame) and reference-based video generation. It is capable of generating videos up to 30 seconds long at 30fps.

Input:

Output:

Input:

—

Output:

$0.03-0.12/second

Context length:

20K

Max output:

—

Alibaba: Wan3.0-Video-Prime is a high-speed version of the Wan3.0 video generation model. It has the same powerful capabilities as the standard Wan3.0-Video model, supports comprehensive reference input of four modalities, can generate videos up to 30 seconds, and significantly improves the end-to-end generation speed while providing an immersive audio-visual experience.

Input:

Output:

Input:

—

Output:

$0.0578-0.238/second

Context length:

20K

Max output:

—

DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, an asymmetric split that keeps per-token compute low relative to the model's total size. Image understanding is native to the architecture, with visual and text embeddings trained jointly from the start of pre-training rather than added afterward as in the earlier experimental V4 Flash Vision Exp(opens in new tab).

Input:

Output:

Input:

$0.15$0.105/1M

Output:

$0.6$0.42/1M

Context length:

1M

Max output:

384K

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding, and long-horizon agent workflows, with strong performance across knowledge, math, and software engineering benchmarks.

Input:

Output:

Input:

$0.66$0.462/1M

Output:

$1.98$1.386/1M

Context length:

1M

Max output:

393K

MoonshotAI: Kimi K3 is Kimi’s most powerful model to date, with 2.8 trillion parameters. It is built based on the hybrid linear attention mechanism Kimi Delta Attention and attention residual technology. It has native visual understanding capabilities and a context window of 1 million tokens. It is suitable for cutting-edge intelligent scenarios such as software engineering, knowledge work, and deep reasoning.

Input:

Output:

Input:

$3$0.9/1M

Output:

$15$4.5/1M

Context length:

1.05M

Max output:

1.05M

Qwen3.8-Flash is the latest multi-modal model of the Qwen series, which combines powerful reasoning and generation capabilities with amazing speed. It natively supports context windows with millions of tokens, and can handle long documents, complete code bases, and complex conversations at once. This model excels in code assistance, agent workflow, and visual understanding, whether it is autonomously repairing code, operating desktop applications, or analyzing charts and long videos. Qwen3.8-Flash is fully compatible with OpenAI and Anthropic API protocols, and can be seamlessly integrated with popular development tools such as Claude Code and Codex, helping developers easily build high-concurrency applications and intelligent workflows. With excellent performance and highly competitive inference costs, Qwen3.8-Flash is an ideal choice for developers and enterprises seeking a balance between performance and cost for AI applications.

Input:

Output:

Input:

$0.16$0.144/1M

Output:

$0.47$0.423/1M

Context length:

1M

Max output:

1M

Qwen: Qwen3.8-Max is the MoE flagship model with 2.4 trillion parameters, achieving a comprehensive leap in the field of programming and professional work. It autonomously codes and delivers complete projects spanning 10+ days, handling hundreds of specialized tasks in legal, finance, design and other professional fields, producing production-grade results end-to-end in a single conversation. Native visual understanding spans the entire lifecycle of planning, execution, and validation, enabling deep semantic analysis of very long documents and extended video content. During long-term tasks, it can plan autonomously, iterate through closed-loop feedback, and continuously evolve.

Input:

Output:

Input:

$2$1.8/1M

Output:

$6$5.4/1M

Context length:

1M

Max output:

64K

Qwen3.8-Max-0902 is a snapshot version of Qwen3.8-Max, which has stronger coding performance and is suitable for complex engineering tasks and long-term independent development. It simultaneously improves the agent's tool usage capabilities and visual understanding capabilities for graph reasoning, document parsing and multi-modal perception, and retains the context window of 1 million tokens, reasoning modes and a complete tool ecosystem.

Input:

Output:

Input:

$2/1M

Output:

$6/1M

Context length:

1M

Max output:

128K

Qwen Image 3 is a unified image generation and editing model from Qwen. It supports precise rendering of text and details as small as 10px, along with a richer world knowledge base than previous generations.

Input:

Output:

Input:

$0.0024/image

Output:

$0.024/image

Context length:

1K

Max output:

—

Qwen Image 3 Pro is an image generation and editing model from Qwen. It supports precise rendering of text and details as small as 10px, along with richer world knowledge compared to previous generations.

Input:

Output:

Input:

$0.0024/image

Output:

$0.032-0.06/image

Context length:

1K

Max output:

—

GLM 5.2 is a large-scale reasoning model from Z.ai. It supports text input and output with a 1M-token context window, and is suited for long-horizon agent workflows, project-level software engineering, and complex multi-step automation.

Input:

Output:

Input:

$1.4$0.42/1M

Output:

$4.4$1.32/1M

Context length:

1M

Max output:

128K

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.

Input:

Output:

Input:

$1.4$0.42/1M

Output:

$4.4$1.32/1M

Context length:

1M

Max output:

128K