Gemini

Gemini models are fast, multimodal, and support large context windows. Thinking is accepted on all models and applied where supported.

Models

Model	Code	Notes
Gemini 3.1 Pro Preview	gemini-3.1-pro-preview	Most capable Gemini 3.1 (preview).
Gemini 3.1 Flash-Lite Preview	gemini-3.1-flash-lite-preview	Lightweight (preview; retiring 2026/7/6).
Gemini 3.1 Flash-Lite	gemini-3.1-flash-lite	Lightweight.
Gemini 3 Flash Preview	gemini-3-flash-preview	Fast (preview).
Gemini 2.5 Pro	gemini-2.5-pro	High capability.
Gemini 2.5 Flash	gemini-2.5-flash	Fast and economical.

Capabilities

• tools, tool_choice, parallel_tool_calls
• thinking / reasoning_effort (ignored on models without thinking support)
• top_k via extra_body

Penalties

frequency_penalty / presence_penalty are not available on preview or Gemini 3 models. See Per-Model Details → Gemini.

Example

from openai import OpenAI

client = OpenAI(base_url="https://llm.siraya.ai/v1", api_key="<API_KEY>")
resp = client.chat.completions.create(
    model="gemini-2.5-flash",
    messages=[{"role": "user", "content": "Describe this image."}],
)
print(resp.choices[0].message.content)