01 · DeepSeek
DeepSeek
- Available model
- DeepSeek-v4.1-Flash
- API model ID
deepseek-flash- Context window
- 1M tokens
- Inputs
- Text; images with V4.1-Flash
- Modes
- Thinking and non-thinking
Models & specifications
Models
Four language model families and three video model families. Actual available models and protocols follow your provisioned catalog.
01 Language & multimodal
Language model families
02 · Zhipu AI
GLM
- Featured model
- GLM-5.3
glm-5.3 - Context window
- 1M tokens, up to 128K output
- Inputs
- Text
- Reasoning
- Always on, with adjustable effort
- Provider
- Zhipu AI (Z.ai)
03 · Alibaba Cloud
Qwen
- Featured model
- Qwen3.8-Max
qwen3.8-max - Version snapshot
- 2 September 2026
qwen3.8-max-0902 - Context window
- 1M tokens
- Inputs
- Text and images
- Provider
- Alibaba Cloud Model Studio
04 · Moonshot AI
Kimi
- Featured model
- Kimi K3.1 Pro
- Context window
- 1M tokens, with context caching
- Inputs
- Text and visual input
- Provider
- Moonshot AI
02 Model specifications
Language model specifications
| Family | Model | Context window | Input |
|---|---|---|---|
| DeepSeek | DeepSeek-v4.1-Flash | 1M tokens | Text; images with V4.1-Flash |
| GLM | GLM-5.3 | 1M tokens, up to 128K output | Text |
| Qwen | Qwen3.8-Max | 1M tokens | Text and images |
| Kimi | Kimi K3.1 Pro | 1M tokens, with context caching | Text and visual input |
03Video creation
Video models
Generate video from text, images, or reference materials.
ByteDance Seed
Seedance 2.5
Generates video with synchronized audio, extended storytelling, and reference-based editing.
- Output
- Video with sound, up to 30 seconds per clip
- Controls
- Edit with references, and make videos longer
Alibaba Tongyi Lab
Wan 3.0
Generates video from text, images, or references, with optional audio.
- Output
- 2 to 30 seconds, 480P to 1080P
- Model IDs
wan3.0-videoandwan3.0-video-prime
MiniMax
MiniMax H3
Generates video from text and images with first- and last-frame control.
- Output
- 768P or 2K, 4 to 15 seconds
- Note
- Earlier MiniMax video models used the Hailuo name; the Hailuo AI app is a separate consumer product.
01
Product demonstrations
Use product photos to illustrate controls and operating steps.
02
Brand films
Develop alternate cuts from a brand palette and a script.
03
Storyboard drafts
Break a script into shots and sketch a visual draft for each one.
04
Video extensions
Add a transition or leave more room for the closing credits.
04 Applications
Common use cases
Documents & office work
Typical tasks
- Summarize meeting notes
- Draft emails and replies
- Extract key points
Document workflowWorkflow overview
Coding & troubleshooting
Typical tasks
- Read and explain code
- Add unit tests and technical documentation
- Draft changes and run checks
Code repair workflowWorkflow overview
Knowledge retrieval & analysis
Typical tasks
- Search internal policies and manuals
- Compare information across documents
- Summarize research findings
Knowledge workflowWorkflow overview
Video generation & storyboarding
Typical tasks
- Generate product demo clips
- Edit visuals using references
- Create storyboard drafts
Video generation workflowWorkflow overview
Prepare a case for a support agent
Typical tasks
- Categorise incoming tickets
- Extract steps to reproduce a fault
- Draft a reply using the returns policy
Customer supportWorkflow overview
Turn unstructured records into importable data
Typical tasks
- Extract quantities from purchase orders
- Normalise field names and dates
- Flag duplicates and missing values
Data preparationWorkflow overview
Keep terminology consistent across languages
Typical tasks
- Translate a manual against a glossary
- Preserve UI placeholders and variables
- Flag ambiguous terms for the translator
LocalisationWorkflow overview
Build practice materials from lesson notes
Typical tasks
- Generate revision cards by chapter
- Draft exercises at several levels
- Prepare presenter notes for training
LearningWorkflow overview
05Sources
Sources
- DeepSeek API changelogDeepSeek-v4.1-Flash updates
- DeepSeek V4 release note1M-token context and thinking modes
- DeepSeek V4.1-Flash release noteNative visual understanding
- Zhipu AI GLM-5.3 model pageFlagship status, context window, and text input
- Alibaba Cloud Model Studio, Qwen3.8-MaxModel ID and the 2 September 2026 snapshot
- Alibaba Cloud Qwen3.8-Max announcementRelease, 1M-token context, and multimodal input
- Kimi K3.1 Pro overviewModel family, inputs, and context
- ByteDance Seed, Seedance 2.5Audio and video generation up to 30 seconds
- Alibaba Cloud Model Studio, video generationWan 3.0 model IDs, durations, and resolutions
- MiniMax models overviewMiniMax H3 capabilities, resolutions, and durations