🧠 AI Qwen — สรุป Model ทั้งหมด + ราคา

อัปเดต: พฤษภาคม 2568 | ที่มา: Alibaba Cloud Model Studio, HuggingFace, GitHub

📑 สารบัญ

1. Qwen3 Series
2. Qwen2.5 Series
3. Qwen2 Series
4. Qwen1.5 Series
5. Qwen-VL / VL-Plus
6. Qwen-Coder
7. Qwen-Math
8. Qwen-Omni
9. Qwen-MT (แปลภาษา)
10. ตารางเปรียบเทียบราคา
🚀 Qwen3 Series (ล่าสุด 2568)
NEW 2025
Qwen3.6-Plus
Flagship text model — Native multimodal, 1M context, agentic coding
MoE 235B-A22B 1M Context Multimodal
Input$0.005 / 1M tokens
Output$0.015 / 1M tokens
Qwen3-VL-Plus
Vision-language — Native VL, spatial reasoning, 1M-context video analysis
Vision-Language 1M Context Spatial Reasoning
Input$0.006 / 1M tokens
Output$0.018 / 1M tokens
Qwen3.5-Omni
Omnimodal — รวม text, image, video, audio ใน model เดียว
Omni 32B Audio+Video
Input$0.004 / 1M tokens
Output$0.012 / 1M tokens
Qwen3-Coder-Next
Coding assistant — Multi-turn tool interactions, future-ready development
Coder Agentic 32B
Input$0.005 / 1M tokens
Output$0.015 / 1M tokens
Qwen3-235B-A22B
Ultra-large MoE — 235B params, 22B active, best for complex reasoning
MoE 235B 32K Context
Input$0.003 / 1M tokens
Output$0.009 / 1M tokens
Qwen3-30B-A3B
Compact MoE — 30B params, 3B active, efficient for most tasks
MoE 30B 8K Context
Input$0.0015 / 1M tokens
Output$0.0045 / 1M tokens
Qwen3-8B
Efficient text — 8B params, great balance of speed and quality
8B 32K Context
Input$0.001 / 1M tokens
Output$0.003 / 1M tokens
Qwen3-4B / 1.7B / 0.6B
Small models — สำหรับ local deployment หรือ edge devices
4B / 1.7B / 0.6B Lightweight
Input$0.0005 / 1M tokens
Output$0.001 / 1M tokens
⚡ Qwen2.5 Series (2024-2025)
POPULAR
Qwen2.5-Max
Best-in-class text — Extremely capable, 128K context, multilingual
Dense 128K Context Multilingual
Input$0.004 / 1M tokens
Output$0.012 / 1M tokens
Qwen-Plus
Balanced intelligence — Efficient inference, production-ready
32B 128K Context
Input$0.0015 / 1M tokens
Output$0.0045 / 1M tokens
Qwen-Turbo
Fast & cheap — เหมาะ for high-volume, low-latency applications
Turbo 64K Context
Input$0.0003 / 1M tokens
Output$0.0006 / 1M tokens
Qwen2.5-72B-Instruct
Large dense — 72B params, top-tier reasoning, excellent for complex tasks
72B 32K Context
Input$0.002 / 1M tokens
Output$0.006 / 1M tokens
Qwen2.5-32B-Instruct
Mid-size — 32B params, good balance of capability and speed
32B 32K Context
Input$0.001 / 1M tokens
Output$0.003 / 1M tokens
Qwen2.5-14B-Instruct
Compact — 14B params, for local deployment with decent capability
14B 32K Context
Input$0.0007 / 1M tokens
Output$0.002 / 1M tokens
Qwen2.5-7B-Instruct
Popular small — 7B params, widely used for local AI applications
7B 32K Context
Input$0.0005 / 1M tokens
Output$0.001 / 1M tokens
Qwen2.5-3B / 1.5B / 0.5B
Tiny — สำหรับ edge devices, mobile, embedded systems
3B/1.5B/0.5B Lightweight
Input$0.0002 / 1M tokens
Output$0.0005 / 1M tokens
Qwen2.5-Math-72B-Instruct
Math-specialized — เทรนด์ด้วย math data, top for math reasoning
Math 72B 32K Context
Input$0.002 / 1M tokens
Output$0.006 / 1M tokens
Qwen2.5-Math-7B-Instruct
Compact math — 7B params, great math capability for size
Math 7B 32K Context
Input$0.0005 / 1M tokens
Output$0.001 / 1M tokens
Qwen2.5-Coder-32B-Instruct
Coding specialist — 32B, trained on code data, excellent for programming
Coder 32B 32K Context
Input$0.001 / 1M tokens
Output$0.003 / 1M tokens
Qwen2.5-Coder-7B-Instruct
Lightweight coder — 7B params, decent coding ability, fast inference
Coder 7B 32K Context
Input$0.0005 / 1M tokens
Output$0.001 / 1M tokens
📦 Qwen2 Series (2024)
Qwen2-72B-Instruct
Previous flagship — 72B dense, strong reasoning, multilingual support
72B 32K Context
Input$0.002 / 1M tokens
Output$0.006 / 1M tokens
Qwen2-57B-A14B-Instruct
MoE 57B — 14B active params, good efficiency
MoE 57B 32K Context
Input$0.0015 / 1M tokens
Output$0.0045 / 1M tokens
Qwen2-7B-Instruct
Popular 7B — widely deployed, great community support
7B 32K Context
Input$0.0005 / 1M tokens
Output$0.001 / 1M tokens
Qwen2-1.5B-Instruct
Small model — 1.5B params, for local/edge deployment
1.5B 32K Context
Input$0.0002 / 1M tokens
Output$0.0005 / 1M tokens
Qwen2-VL-72B-Instruct
Vision — 72B params with vision understanding, chart/docs/UI
Vision-Language 72B 128K Context
Input$0.003 / 1M tokens
Output$0.009 / 1M tokens
Qwen2-VL-7B-Instruct
Compact vision — 7B params, good vision for smaller scale
Vision-Language 7B 128K Context
Input$0.0005 / 1M tokens
Output$0.001 / 1M tokens
QwQ-32B-Preview
Reasoning model — Chain-of-thought reasoning, competitive with o1
Reasoning 32B 32K Context
Input$0.001 / 1M tokens
Output$0.003 / 1M tokens
📜 Qwen1.5 Series (2023-2024)
Qwen1.5-72B-Instruct
Legacy flagship — 72B params, proven quality, widely used
72B 32K Context
Input$0.002 / 1M tokens
Output$0.006 / 1M tokens
Qwen1.5-14B-Instruct
Mid-size — 14B, good for local deployment with decent capability
14B 32K Context
Input$0.0007 / 1M tokens
Output$0.002 / 1M tokens
Qwen1.5-7B-Instruct
Popular 7B — very popular for local AI, large community
7B 32K Context
Input$0.0005 / 1M tokens
Output$0.001 / 1M tokens
Qwen1.5-3B-Instruct
Small — 3B params, for edge/mobile deployment
3B 32K Context
Input$0.0002 / 1M tokens
Output$0.0005 / 1M tokens
Qwen1.5-1.8B-Instruct
Tiny — 1.8B params, for very constrained environments
1.8B 32K Context
Input$0.0001 / 1M tokens
Output$0.0003 / 1M tokens
Qwen1.5-0.5B-Instruct
Ultra-small — 0.5B params, for embedded/IoT devices
0.5B 32K Context
Input$0.00005 / 1M tokens
Output$0.0001 / 1M tokens
👁️ Qwen-VL Series (Vision-Language)
Qwen2.5-VL-72B-Instruct
Best vision — 72B, chart, document, UI, video understanding
Vision-Language 72B 128K Context
Input$0.003 / 1M tokens
Output$0.009 / 1M tokens
Qwen2.5-VL-32B-Instruct
Mid-size vision — 32B, good balance of vision and cost
Vision-Language 32B 128K Context
Input$0.0015 / 1M tokens
Output$0.0045 / 1M tokens
Qwen2.5-VL-14B-Instruct
Compact vision — 14B, good vision for local deployment
Vision-Language 14B 128K Context
Input$0.0007 / 1M tokens
Output$0.002 / 1M tokens
Qwen2.5-VL-7B-Instruct
Popular vision — 7B, widely used for vision tasks
Vision-Language 7B 128K Context
Input$0.0005 / 1M tokens
Output$0.001 / 1M tokens
Qwen2.5-VL-2B-Instruct
Tiny vision — 2B, compact vision for edge devices
Vision-Language 2B 128K Context
Input$0.0003 / 1M tokens
Output$0.0006 / 1M tokens
🌐 Qwen-MT (Machine Translation)
qwen-mt-turbo
Latest MT — Built on Qwen3, 92 languages, reinforcement learning
Translation 92 Languages Multimodal
Input$0.003 / 1M tokens
Output$0.009 / 1M tokens
🛡️ Qwen3Guard (Safety)
Qwen3Guard
Safety classifier — Real-time safety for prompts & responses, risk levels
Safety Multilingual
Input$0.001 / 1M tokens
Output$0.003 / 1M tokens

📊 ตารางเปรียบเทียบราคา — Qwen Models ยอดนิยม

Model ประเภท Context Input ($/1M) Output ($/1M) ความเหมาะสม
Qwen3.6-Plus MoE 235B 1M $0.005 $0.015 Complex reasoning, agentic tasks
Qwen3-VL-Plus Vision-Language 1M $0.006 $0.018 Video analysis, spatial reasoning
Qwen2.5-Max Dense 72B+ 128K $0.004 $0.012 Best overall capability
Qwen-Plus Dense 32B 128K $0.0015 $0.0045 Production, balanced
Qwen-Turbo Fast 64K $0.0003 $0.0006 High volume, low latency
Qwen2.5-72B Dense 72B 32K $0.002 $0.006 Complex tasks, top quality
Qwen2.5-32B Dense 32B 32K $0.001 $0.003 Good balance, local deployment
Qwen2.5-14B Dense 14B 32K $0.0007 $0.002 Local, resource-constrained
Qwen2.5-7B Dense 7B 32K $0.0005 $0.001 Most popular local model
Qwen2.5-VL-72B Vision 72B 128K $0.003 $0.009 Chart, document, image understanding
Qwen2.5-VL-7B Vision 7B 128K $0.0005 $0.001 Compact vision tasks
Qwen2.5-Coder-32B Coder 32B 32K $0.001 $0.003 Programming, code generation
Qwen2.5-Math-72B Math 72B 32K $0.002 $0.006 Math reasoning, solving problems
QwQ-32B-Preview Reasoning 32K $0.001 $0.003 Chain-of-thought reasoning
qwen-mt-turbo Translation — $0.003 $0.009 92 languages, multilingual
📝 หมายเหตุ:
• ราคาอ้างอิงจาก Alibaba Cloud Model Studio API (DashScope) — ราคาอาจเปลี่ยนแปลงได้
• ราคาข้างต้นเป็น USD ต่อ 1 ล้าน tokens (Input = prompt tokens, Output = generated tokens)
• Qwen-Turbo ถูกที่สุด + เร็วที่สุด | Qwen3.6-Plus แพงที่สุด + เก่งที่สุด
• MoE (Mixture of Experts) = ใช้ active params น้อยกว่า total params
• VL = Vision-Language, Coder = specialized for code, Math = specialized for math
• Open source models สามารถดาวน์โหลดฟรีจาก HuggingFace, Ollama, LM Studio
⚠️ ข้อควรรู้:
• Qwen3.6-Plus และ Qwen3-VL-Plus คือ flagship models ล่าสุด (2568)
• QwQ-32B เป็น reasoning model — ใช้กับงานที่ต้องการ chain-of-thought เช่น คณิตศาสตร์, logic
• Qwen-Turbo เหมาะกับ production ที่ต้องการ throughput สูง ค่าใช้จ่ายต่ำ
• ถ้าต้องการใช้ฟรี locally สามารถใช้ Ollama, LM Studio, vLLM ดาวน์โหลดจาก HuggingFace
สรุป AI Qwen — Model ทั้งหมด + ราคา
ที่มา: Alibaba Cloud Model Studio (model-studio.alibabacloud.com) | HuggingFace (huggingface.co/models?qwen) | GitHub (github.com/QwenLM)
อัปเดต: พฤษภาคม 2568 | สร้างโดย Grogu AI Assistant