Z.AI MODEL SELECTION GUIDE · 2026
EN

Z.ai Model Selection Guide

Zhipu Z.ai full lineup · text generation · multimodal understanding · image/video/voice · vector retrieval — Sources: BigModel model overview · Pricing (synced 2026-08-30)

69+Models
8Categories
1MMax context
Tier
FlagshipBalancedBudget
Speed
FastestFastStandardSlowSlowest
Modality
TextImageAudioVideoCode
Reasoning
BestDeepStandardBasicFast
Price
Premium¥20+ / ¥72+Expensive¥8-20 / ¥36-72Costly¥2-8 / ¥8-36Moderate¥0.5-2 / ¥2-8Affordable¥0.1-0.5 / ¥0.4-2Cheapest<¥0.1 / <¥0.4Unit: CNY / 1M tokens (input / output), at official list price

Quick Picks

Direct recommendation by task type

Chat / WritingGLM-5.3
Complex reasoning / long-horizon AgentGLM-5.3
Coding / software engineeringGLM-5.3
Image / video understandingGLM-5.3-Flash
CheapestGLM-4.7-Flash
Long-document processingGLM-4-Long
Image generationGLM-Image
Video generationCogVideoX-3
Realtime voice chatGLM-Realtime-Flash
Speech recognitionGLM-ASR-2512
Speech synthesisGLM-TTS
Semantic search / RAGEmbedding-3

Naming Conventions

Understand these patterns and instantly grasp any model name

ElementMeaningExample
GLM prefix Zhipu general-purpose LLM family GLM-5.3, GLM-4.7
Version number (4→5→5.3) Higher number = newer and stronger GLM-4 → GLM-5 → GLM-5.3
-Flash Lightweight and fast, budget-friendly GLM-5.3-Flash, GLM-4.7-Flash
-FlashX Enhanced high-speed Flash GLM-4.7-FlashX
-Turbo Speed-optimized version GLM-5-Turbo, GLM-5V-Turbo
-Air Lightweight, balanced GLM-4.5-Air
-AirX Ultra-fast, low-latency GLM-4.5-AirX
-Long Built for ultra-long context GLM-4-Long
-V Vision understanding model GLM-5V-Turbo, GLM-4.6V
-V-Thinking Vision reasoning model GLM-4.1V-Thinking-FlashX
GLM-Realtime Realtime audio-video conversation GLM-Realtime
GLM-ASR Automatic speech recognition GLM-ASR-2512
GLM-TTS Text to speech GLM-TTS
CogView Image generation CogView-4
CogVideoX Video generation CogVideoX-3
Vidu Video generation (partner) Vidu Q1
Embedding Text embedding Embedding-3
-YYYYMMDD Date snapshot GLM-4-Flash-250414

Frontier Flagship Models

Zhipu's latest flagship generation, recommended for complex reasoning, coding, agents and multimodal tasks

Model IDTierPriceModalityReasoningSpeedContextInputOutputNotes
GLM-5.3NEW Flagship
Best Standard 1M 1M 128K Latest flagship for complex software engineering and long-horizon agent tasks; coding experience 50% better than the previous generation
GLM-5.3-FlashNEW Balanced
Deep Fast 1M 1M 128K Native multimodal budget model; natively understands images/videos/files; list price ¥0.8/¥2.8, limited-time 50% off (¥0.4/¥1.4)
GLM-5.2 Flagship
Best Standard 1M 1M 128K Stable execution of complex long-horizon tasks; significantly improved coding
GLM-5.1 Flagship
Deep Standard 200K 200K 128K Coding on par with Claude Opus 4.6; autonomous long-horizon tasks for up to 8 hours; ¥8/¥28 when input ≥32K
GLM-5 Flagship
Deep Standard 200K 200K 128K Coding on par with Claude Opus 4.5; excels at agentic long-horizon planning and execution; ¥6/¥22 when input ≥32K
GLM-5V-Turbo Flagship
Deep Fast 200K 200K 128K Multimodal coding base, balancing vision understanding, reasoning and code generation; ¥7/¥26 when input ≥32K
GLM-5-Turbo Balanced
Deep Fast 200K 200K 128K Optimized for long-task core needs; good continuity on complex long tasks; ¥7/¥26 when input ≥32K

Text Models

Text models focused on natural-language understanding and generation, covering chat, writing, reasoning and code

Model IDTierPriceModalityReasoningSpeedContextInputOutputNotes
GLM-4.7 Balanced
Standard Standard 200K 200K 128K Upgraded general chat, reasoning and agent capabilities; ¥2/¥8 when input <32K and output <0.2K
GLM-4.7-FlashX Budget
Basic Fastest 200K 200K 128K Lightweight high-speed Flash; fast reasoning; ideal for high-concurrency calls; ¥0.5 in / ¥3 out
GLM-4.7-Flash Budget
Basic Fastest 200K 200K 128K Free text model carrying over the general capabilities of the GLM-4.7 base
GLM-4.6 Balanced
Standard Standard 200K 200K 128K Excels at advanced coding, complex reasoning and tool calling; pay-as-you-go pricing not public; deployable via private instances
GLM-4.5-Air Balanced
Standard Standard 128K 128K 96K Lightweight model; stable on reasoning, coding and agent tasks
GLM-4.5-AirX Budget
Basic Fast 128K 128K 96K Ultra-fast version for low-latency, high-response business scenarios
GLM-4-Long Balanced
Standard Standard 1M 1M 4K Designed for ultra-long text and memory-intensive tasks
GLM-4-FlashX-250414 Budget
Basic Fastest 128K 128K 16K Enhanced high-speed Flash; ideal for high-concurrency calls
GLM-4.5-Flash Budget
Basic Fastest 128K 128K 96K Free text model; handles up to 128K context
GLM-4-Flash-250414 Budget
Basic Fastest 128K 128K 16K Free text model
GLM-4-Plus Flagship
Deep Standard 128K 128K 128K High-intelligence flagship; fully self-developed 4th-gen base model; supports advanced agent capabilities
GLM-4-Air-250414 Balanced
Standard Standard 128K 128K 128K GLM-4-Air date snapshot 250414; cost-effective text model
GLM-4-AirX Balanced
Standard Fast 8K 8K 8K Ultra-fast reasoning version for low-latency, high-response scenarios
GLM-4-Assistant Balanced
Standard Standard 128K 128K 128K Text model optimized for agent applications
GLM-Z1-Air Balanced
Deep Standard 128K 128K 128K Cost-effective reasoning model
GLM-Z1-AirX Balanced
Deep Fast 32K 32K 32K Ultra-fast reasoning model, balancing depth and response speed
GLM-Z1-FlashX Budget
Standard Fastest 128K 128K 16K High-speed, low-cost reasoning model for high-concurrency reasoning
GLM-Z1-Flash Budget
Standard Fastest 128K 128K 16K Free reasoning model
GLM-4-9B Balanced
Basic Standard 128K 128K 128K Open-source 9B text model
ChatGLM3-6B Balanced
Basic Standard 8K 8K 8K Open-source chat model (3rd gen)

Vision understanding model

Image and video understanding, vision reasoning and mobile agent models

Model IDTierPriceModalityReasoningSpeedContextInputOutputNotes
GLM-4.6V Balanced
Standard Standard 128K 128K 32K Native tool calling; stable frontend code replication; ¥1/¥3 when input <32K, ¥2/¥6 for [32,128)
GLM-4.5V Balanced
Standard Standard 64K 64K 32K Vision understanding model; ¥2/¥6 when input <32K
GLM-4V-Plus-0111 Balanced
Standard Standard 16K 16K Vision reasoning model, ¥4/1M tokens
AutoGLM-Phone Flagship
Standard Standard 20K 20K 2048 Mobile AI assistant framework; completes app operations via natural language
GLM-4.1V-Thinking-FlashX Balanced
Deep Fast 64K 64K 16K Excels at complex-scene understanding and multi-step analysis; ideal for high-concurrency vision reasoning
GLM-OCR Balanced ¥0.2 / 1M tokens
Basic Fast 32K 32K Lightweight image-text parsing model balancing high-accuracy, high-efficiency document understanding; input: single image ≤10MB / PDF ≤50MB / 100 pages
GLM-4.6V-FlashX Budget
Basic Fastest 128K 128K 32K High-speed vision reasoning model; ¥0.15/¥1.5 when input <32K, ¥0.3/¥3 for [32,128)
GLM-4.6V-Flash Budget
Basic Fastest 128K 128K 32K Free model with vision reasoning
GLM-4.1V-Thinking-Flash Budget
Standard Fastest 64K 64K 16K Free model with vision reasoning
GLM-4V-Flash Budget
Fast Fastest 16K 16K 1K Free model with image understanding

Image Generation Models

Text-to-image and image editing models, billed per request

Model IDTierPriceModalityNotes
GLM-Image Flagship ¥0.1 / request
Flagship image generation model; stronger complex-instruction following and knowledge-dense generation; outstanding text rendering; multi-resolution
CogView-4 Balanced ¥0.06 / request
General image generation model; high quality with richer detail; suits diverse creative scenarios; multi-resolution
CogView-3-Flash Budget Free
Free model for lightweight image creation; multi-resolution

Video Generation Models

Text-to-video, image-to-video and first/last-frame generation models, billed per request

Model IDTierPriceModalityNotes
CogVideoX-3 Flagship ¥1 / request
Flagship video model; stronger instruction following and physics simulation; improved realism and 3D scenes; supports first/last-frame generation
ViduQ1-Text Balanced ¥2.5 / request
Vidu Q1 text-to-video, 1080p
ViduQ1-Image Balanced ¥2.5 / request
Vidu Q1 image-to-video, 1080p
ViduQ1-Start-End Balanced ¥2.5 / request
Vidu Q1 first/last-frame generation, 1080p
Vidu2-Image Balanced ¥1.25 / request
Vidu 2 image-to-video, 720p
Vidu2-Start-End Balanced ¥1.25 / request
Vidu 2 first/last-frame generation, 720p
Vidu2-Reference Balanced ¥2.5 / request
Vidu 2 reference-to-video, 720p
CogVideoX-2 Balanced ¥0.5 / request
General video generation model; multi-resolution
CogVideoX-Flash Budget Free
Free video generation model; accepts image and text input

Audio & Video Models

Speech recognition, speech synthesis, realtime voice/video conversation and voice cloning models, with varying billing units

Model IDTierPriceModalityNotes
GLM-Realtime-Flash Flagship Audio ¥0.18/min · Video ¥1.2/min
Realtime audio-video model, low-latency version; accepts video, audio and text multimodal input
GLM-Realtime-Air Balanced Audio ¥0.3/min · Video ¥2.1/min
Realtime audio-video model, lightweight balanced version; accepts video, audio and text multimodal input
GLM-4-Voice Balanced ¥80 / 1M tokens
Realtime voice chat model; supports text and audio; audio token unit price, not directly comparable to text tiers
GLM-TTS-Clone Balanced ¥6 / request
Voice cloning model; 3 seconds of audio is enough to clone a similar voice
GLM-TTS Balanced ¥2 / 10K chars
Speech synthesis model; ultra-humanlike voice with emotion; offers streaming and non-streaming APIs
GLM-ASR-2512 Balanced Input ¥16/1M tokens (~¥0.0002/sec), free output
High-accuracy speech recognition; low character error rate; custom vocabulary; covers major languages and dialects

Embedding & Others

Specialized models for vector retrieval, reranking, roleplay, emotional support and code completion

Model IDTierPriceModalityContextOutputNotes
Embedding-3 Balanced
8K 3rd-gen text embedding model (V3) for semantic retrieval, clustering, topic modeling and classification
Embedding-2 Balanced
8K 2nd-gen text embedding model (V2)
Rerank Balanced
4K Text reranking model, ¥0.8/1M input tokens
CodeGeeX-4 Balanced
128K 32K Code completion model
CharGLM-4 Balanced
8K 4K Humanlike chat model
Emohaa Balanced
8K 4K Emotional support model

Legacy Models

Older models no longer recommended or under maintenance; for compatibility reference only

Model IDTierPriceModalityReasoningSpeedContextInputOutputNotes
GLM-4-0520 Balanced
Standard Standard 128K 128K 128K GLM-4 0520 legacy version
GLM-4V-Plus Balanced
Standard Standard 8K 8K 8K GLM-4V Plus legacy version
GLM-4V Balanced
Standard Standard 8K 8K 8K GLM-4V legacy version
GLM-4 Balanced
Standard Standard 128K 128K 128K GLM-4 base legacy version
GLM-4-Air Balanced
Standard Standard 128K 128K 128K GLM-4-Air legacy version
GLM-4-Flash Budget
Basic Fastest 128K 128K 16K GLM-4-Flash legacy version, free
GLM-4.5 Balanced
Standard Standard 128K 128K 128K Pay-as-you-go API retired; available via fine-tuning and private deployment
CogView-3 Balanced
Basic Standard Pay-as-you-go API retired; private deployment only; historical price ¥0.06/request

Capability Matrix

Find the primary and alternative model for each typical scenario

ScenarioRecommendedAlternativesKey capability
Complex coding / software engineering GLM-5.3 GLM-5.2 Reasoning Long context Agent
Deep reasoning / math derivation GLM-5.3 GLM-5.1 Best reasoning 128K output
Long-document processing GLM-4-Long GLM-5.3 1M context Low cost
Image / video understanding GLM-5.3-Flash GLM-4.6V Native multimodal Video understanding
Image generation GLM-Image CogView-4 Text rendering Multi-resolution
Video generation CogVideoX-3 ViduQ1-Text First/last frame Physics simulation
Realtime voice chat GLM-Realtime-Flash GLM-4-Voice Realtime Audio-video
Speech recognition GLM-ASR-2512 High accuracy Dialects
Semantic search / RAG Embedding-3 Embedding-2 8K context Semantic retrieval
Cheapest GLM-4.7-Flash GLM-4-Flash-250414 Free 200K context