Documentation
Model-by-Model Differences
International Models
| Model | Preferred Format | Context | Reasoning | Key Notes |
|---|---|---|---|---|
| Claude (Anthropic) | XML | 200K+ | No (has extended thinking) | Wrap each semantic block in its own <tag>; also produce SKILL.md (progressive disclosure) |
| GPT non-reasoning (4o/5) | Markdown | 128K+ | No | High few-shot payoff; use Structured Outputs for JSON |
| o-series / GPT-5-Thinking | Plain | 200K+ | Yes | Strip CoT prompts; currently no system role → merged into first user turn |
| Gemini 1.5/2.0/2.5 | Markdown | 1M-2M | No | Native long-context — inline knowledge directly; function_declarations types must be UPPERCASE |
Chinese Models
| Model | Preferred Format | Context | Reasoning | Key Notes |
|---|---|---|---|---|
| Qwen 2.5 / 3 | Markdown | 128K | No | OpenAI-compatible; native Chinese; Bailian agent/workflow support |
| QwQ / Qwen3-Thinking | Plain | 128K | Yes | Aligns with o1/R1 style; no CoT prompts |
| DeepSeek V3 | Markdown | 128K | No | OpenAI-compatible API; strong on code & math |
| DeepSeek R1 | Plain | 128K | Yes | Auto-produces <think>; never add 'think step by step' |
| Kimi (Moonshot) | Natural Chinese | 128K-2M | No | Huge context; paste long docs straight into system |
| Zhipu GLM-4 | Markdown | 128K | No | Strong Chinese writing; OpenAI-compatible |
| ERNIE (Baidu) | Markdown | 128K | No | Best for China-local context; Qianfan prefers workflow DSL |
| Doubao / Coze (ByteDance) | Markdown + DSL | 128K | No | Fast & cheap; Coze = persona + plugins + KB + workflow |
NOTE
PromptMan's adapters encode these characteristics — you don't need to memorize them manually.
Continue Reading