Model generasi teks - Qwen, DeepSeek, GLM | Model Studio - Alibaba Cloud Model Studio

Pilih model generasi teks yang tepat untuk Agen AI, chatbot, dan pemrosesan dokumen.

Model yang direkomendasikan untuk tool coding

Kami merekomendasikan qwen3.7-plus karena menawarkan keseimbangan optimal antara performa dan biaya, dengan dukungan penuh tool calling serta jendela konteks 1 juta token untuk codebase besar. Untuk kemampuan reasoning terkuat, pilih qwen3.7-max.

Migrasi dari model closed-source

Petakan model GPT, Claude, atau Gemini Anda saat ini ke model Bailian yang setara.

	Contoh closed-source	Rekomendasi Bailian
Kemampuan tertinggi	GPT-5.5, Claude Opus 4.7, Gemini 3.1 Pro	`qwen3.7-max`
Seimbang	GPT-5.4, Claude Sonnet 4.6, Gemini 3 Pro	`qwen3.7-plus`, `deepseek-v4-pro`, `glm-5.2`
Ringan & berbiaya rendah	GPT-5.4-mini, Claude Haiku 4.5, Gemini 3.1 Flash	`qwen3.6-flash`, `deepseek-v4-flash`, `MiniMax-M2.5`

Kasus penggunaan

Mulai dengan qwen3.7-plus untuk chatbot, Pembuatan konten, Ringkasan, dan pemrosesan dokumen—model ini menyeimbangkan performa, biaya, dan tool bawaan dengan jendela konteks 1 juta token. Untuk mengurangi biaya, beralihlah ke qwen3.6-flash, yang menawarkan kemampuan serupa dengan harga lebih rendah. Untuk reasoning terkuat, gunakan qwen3.7-max (konteks 1 juta token, biaya lebih tinggi).

Jendela konteks

1 juta token kira-kira setara dengan 750.000 kata dalam bahasa Inggris, atau 8–10 novel.

Untuk dokumen panjang atau codebase besar: qwen3.7-plus / qwen3.6-flash (1 juta token).
Untuk tugas standar: 128k–256k token biasanya sudah cukup.

Untuk detail jendela konteks, kunjungi halaman Models.

Tiongkok (Beijing) | Singapura | AS | Tiongkok (Hong Kong) | Frankfurt

Thinking mode

Lakukan reasoning langkah demi langkah untuk matematika multi-langkah, debugging kode, perencanaan arsitektur, dan cross-referensi hukum.

Aktifkan dengan parameter enable_thinking, atau gunakan reasoning.effort pada Responses API untuk mengatur kedalaman thinking. Semua model Qwen3+ mendukung fitur ini, sebagian besar beroperasi dalam mode hybrid yang dapat diaktifkan per permintaan.

Lihat Deep thinking.

Function calling dan built-in tools

Izinkan model melakukan aksi seperti menanyakan cuaca, mencari database, atau memesan rapat.

Function calling (tool kustom yang dipanggil model): Didukung oleh semua model tujuan umum.
Built-in tools (web search, code interpreter, web scraping) tanpa perlu konfigurasi tambahan.

Lihat Tool calling.

Structured output

Paksa output berupa JSON valid, berguna untuk mengekstraksi data terstruktur seperti nama dan alamat dari teks.

Lihat Structured output.

Batch inference

Proses volume besar dengan biaya lebih rendah ketika latensi tidak menjadi prioritas utama.

Lihat Batch inference.

Model yang direkomendasikan

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen3.7-max`	1M	Supported	Supported	Supported	Supported	Supported
`qwen3.7-plus`	1M	Supported	Supported	Supported	Supported	Supported
`qwen3.6-flash`	1M	Supported	Supported	Supported	Supported	Supported
`kimi-k2.6`	256k	Supported	Supported	Unsupported	Unsupported	Unsupported
`deepseek-v4-pro`	1M	Supported	Supported	Unsupported	Unsupported	Unsupported
`deepseek-v4-flash`	1M	Supported	Supported	Unsupported	Unsupported	Unsupported
`glm-5.2`	198k	Supported	Supported	Unsupported	Supported	Unsupported
`MiniMax-M2.5`	192k	Supported	Supported	Unsupported	Unsupported	Unsupported

Model lawas

Untuk proyek baru, gunakan seri Qwen3.6 atau Qwen3.5. Model berikut merupakan model lawas dan tidak lagi direkomendasikan. Kunjungi halaman Models untuk melihat parameter model secara rinci, seperti jendela konteks dan penagihan.

Qwen3.6

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen3.6-max-preview`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3.6-plus`	1M	Supported	Supported	Supported	Supported	Supported

Qwen3.5

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen3.5-plus`	1M	Supported	Supported	Supported	Supported	Supported
`qwen3.5-flash`	1M	Supported	Supported	Supported	Supported	Supported
`qwen3.5-397b-a17b`	256k	Supported	Supported	Supported	Supported	Unsupported
`qwen3.5-122b-a10b`	256k	Supported	Supported	Supported	Supported	Unsupported
`qwen3.5-27b`	256k	Supported	Supported	Supported	Supported	Unsupported
`qwen3.5-35b-a3b`	256k	Supported	Supported	Supported	Supported	Unsupported

Qwen3

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen3-max` View snapshots `qwen3-max-2026-01-23` `qwen3-max-preview` `qwen3-max-2025-09-23`	256k	Supported	Supported	Supported	Supported	Supported
`qwen3-235b-a22b`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-235b-a22b-thinking-2507`	256k	Supported	Supported	Unsupported	Unsupported	Unsupported
`qwen3-235b-a22b-instruct-2507`	256k	Unsupported	Supported	Unsupported	Supported	Unsupported
`qwen3-next-80b-a3b-thinking`	256k	Supported	Supported	Unsupported	Unsupported	Unsupported
`qwen3-next-80b-a3b-instruct`	256k	Unsupported	Supported	Unsupported	Supported	Unsupported
`qwen3-32b`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-30b-a3b`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-30b-a3b-thinking-2507`	256k	Supported	Supported	Unsupported	Unsupported	Unsupported
`qwen3-30b-a3b-instruct-2507`	256k	Unsupported	Supported	Unsupported	Supported	Unsupported
`qwen3-14b`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-8b`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-4b`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-1.7b`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-0.6b`	256k	Supported	Supported	Unsupported	Supported	Unsupported

Qwen3-Coder

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen3-coder-plus` View snapshots `qwen3-coder-plus-2025-09-23` `qwen3-coder-plus-2025-07-22`	1M	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-coder-flash` View snapshots `qwen3-coder-flash-2025-07-28`	1M	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-coder-next`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`qwen3-coder-480b-a35b-instruct`	256k	Unsupported	Supported	Unsupported	Supported	Unsupported
`qwen3-coder-30b-a3b-instruct`	256k	Unsupported	Supported	Unsupported	Supported	Unsupported

Qwen2.5 (open source)

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen2.5-omni-7b`	1M	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen2.5-vl-72b-instruct`	1M	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen2.5-vl-32b-instruct`	1M	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen2.5-vl-7b-instruct`	1M	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen2.5-vl-3b-instruct`	1M	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen2.5-72b-instruct`	1M	Unsupported	Supported	Unsupported	Supported	Unsupported
`qwen2.5-32b-instruct`	1M	Unsupported	Supported	Unsupported	Supported	Unsupported
`qwen2.5-14b-instruct`	1M	Unsupported	Supported	Unsupported	Supported	Unsupported
`qwen2.5-14b-instruct-1m`	1M	Unsupported	Supported	Unsupported	Supported	Unsupported
`qwen2.5-7b-instruct`	1M	Unsupported	Supported	Unsupported	Supported	Unsupported
`qwen2.5-7b-instruct-1m`	1M	Unsupported	Supported	Unsupported	Supported	Unsupported

Translation

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen-mt-plus`	16k	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen-mt-turbo`	16k	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen-mt-flash`	16k	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen-mt-lite`	16k	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported

Qwen-Long

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen-long` View snapshots `qwen-long-2025-01-25`	10M	Unsupported	Unsupported	Unsupported	Supported	Supported
`qwen-long-latest`	10M	Unsupported	Unsupported	Unsupported	Supported	Supported

Role-playing

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen-plus-character`	32k	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen-plus-character-ja`	32k	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen-flash-character`	8k	Unsupported	Unsupported	Unsupported	Unsupported	Unsupported

Legacy Qwen

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`qwen-plus` and its snapshots	1M	Supported	Supported	Supported	Supported	Supported (mainline version only)
`qwen-max` and its snapshots	128k	Unsupported	Supported	Supported	Supported	Supported (mainline version only)
`qwen-flash` and its snapshots	1M	Supported	Supported	Supported	Supported	Supported (mainline version only)
`qwen-turbo` and its snapshots	1M	Supported	Supported	Supported	Supported	Supported (mainline version only)
`qwq-plus`	128k	Supported	Supported	Unsupported	Unsupported	Supported
`qvq-max` and its snapshots	128k	Supported	Unsupported	Unsupported	Unsupported	Unsupported
`qwen-omni-turbo` and its snapshots	32k	Unsupported	Unsupported	Unsupported	Unsupported	Supported (mainline version only)

Third-party models

Model ID	Context	Thinking mode	Function Calling	Built-in tools	Structured output	Batch calling
`glm-5.1`	198k	Supported	Supported	Unsupported	Supported	Unsupported
`glm-5`	198k	Supported	Supported	Unsupported	Supported	Unsupported
`glm-4.7`	198k	Supported	Supported	Unsupported	Supported	Unsupported
`glm-4.5`	198k	Supported	Supported	Unsupported	Supported	Unsupported
`glm-4.5-air`	198k	Supported	Supported	Unsupported	Supported	Unsupported
`MiniMax-M2.7`	200k	Supported	Supported	Unsupported	Unsupported	Unsupported
`MiniMax-M2.1`	200k	Supported	Supported	Unsupported	Unsupported	Unsupported
`kimi-k2.5`	256k	Supported	Supported	Unsupported	Unsupported	Unsupported
`kimi-k2-thinking`	256k	Supported	Supported	Unsupported	Supported	Unsupported
`Moonshot-Kimi-K2-Instruct`	256k	Unsupported	Supported	Unsupported	Unsupported	Unsupported
`deepseek-v3.2`	128k	Supported	Supported	Unsupported	Unsupported	Supported
`deepseek-v3.2-exp`	128k	Supported	Supported	Unsupported	Unsupported	Unsupported
`deepseek-v3.1`	128k	Supported	Supported	Unsupported	Unsupported	Unsupported
`deepseek-v3`	128k	Unsupported	Supported	Unsupported	Unsupported	Supported
`deepseek-r1`	128k	Supported	Supported	Unsupported	Unsupported	Supported
`deepseek-r1-0528`	128k	Supported	Supported	Unsupported	Unsupported	Unsupported
`deepseek-r1-distill-llama-70b`	128k	Supported	Unsupported	Unsupported	Unsupported	Unsupported
`deepseek-r1-distill-qwen-32b`	128k	Supported	Unsupported	Unsupported	Unsupported	Unsupported
`deepseek-r1-distill-qwen-14b`	128k	Supported	Unsupported	Unsupported	Unsupported	Unsupported
`deepseek-r1-distill-qwen-7b`	128k	Supported	Unsupported	Unsupported	Unsupported	Unsupported
`deepseek-r1-distill-qwen-1.5b`	128k	Supported	Unsupported	Unsupported	Unsupported	Unsupported
`deepseek-r1-distill-llama-8b`	128k	Supported	Unsupported	Unsupported	Unsupported	Unsupported