Context windows and maximum output tokens across leading models

Machine-assisted translation of the archived article. Read the Chinese original.

The author argues that increasing the output limit from 4K to 8K tokens can increase compute costs sharply.
Context windows and maximum output tokens for major language models
No.ModelProviderContext windowmax output tokenNotes
1o1 and variantsopenAI200k100kThe author’s top pick at publication
2o1-mini and variantsOpenAi128k64k
3o1-preview and variantsopenai128k32k
4GPT-4o and variantsopenai128k4k
5GPT-4 and variantsopenai8-32k4k
6GPT-4o mini and variantsopenai128k4k
7Claude 3.5 and variantsopenai200k4k
8Moonshot-v1-32KMoonshot AI32k4k
9Moonshot-v1-128KMoonshot AI128k4k
10deepseek v3DeepSeek64k8k
11Doubao-proByteDance256k4k
12Doubao-lite-32kByteDance32k4k
13Doubao-lite-128kByteDance128k4k
14GLM-4 and variantsZhipu AI8k4k
15GLM-Zero-Preview Zhipu AI16k12kOf the 16K total tokens, this model allows up to 12K tokens of output.
16GLM-4-PlusZhipu AI128K4k
17hunyuan-large-longcontextTencent Hunyuan128k6k
18hunyuan-largeTencent Hunyuan28k4k
19hunyuan-turbo and variantsTencent Hunyuan28k4k
20Qwen-TurboAlibaba1000k4k
21Qwen-LongAlibaba1000k4k
22Qwen-PlusAlibaba128k4k
23Qwen-MaxAlibaba32k2k