Quick reference
| Mode | Minimum | Recommended | Maximum |
|---|
| Standard | — | 4096 | 32768 |
| Thinking | 16000 | 16384 | 32768 |
Standard mode
response = client.chat.completions.create(
model="orchid01",
messages=[...],
max_tokens=4096, # default
)
8192.
Thinking mode
max_tokens below 16,000 in thinking mode risks truncated responses. Reasoning tokens count toward the limit — the model reasons before it answers, consuming tokens before the final response begins.
response = client.chat.completions.create(
model="orchid01",
messages=[...],
max_tokens=16384, # minimum for thinking mode
extra_body={"orchid": {"thinking": True}},
)
max_tokens below 16,000 with thinking enabled, Orchid raises it automatically to 16,000 and adds max_tokens_adjusted: true to the response orchid field.
Recommendations by task
| Task | Mode | max_tokens |
|---|
| Extract a specific figure | Standard | 1024 |
| Summarise a filing section | Standard | 2048 |
| Full document analysis | Standard | 4096–8192 |
| Covenant extraction from long agreement | Standard | 8192 |
| Complex multi-document analysis | Thinking | 16384 |
| Multi-step scenario modelling | Thinking | 32768 |
