Models & limits

What you can call, how much, and which parameters work.

Models

Fact	Qwen3.5 9B
Model id	`qwen3.5-9b`
Vendor	Qwen (Alibaba)
Parameters	9B
Quantization	Q4_K_M
Context window	16,384 tokens (prompt + reply)
Max output	2,048 tokens
Input	Text
Capabilities	Chat, Streaming, Tool calling, JSON mode
License	Apache-2.0
Price	Free during beta
Hosting	Spain (EU)

GET /v1/models returns the same list as JSON.

Beta limits apply per API key.

Limit	Value	When you hit it
Requests per minute	60	`429` rate_limit_requests
Tokens per day	500,000	`429` rate_limit_tokens. Resets at 00:00 UTC.
Concurrent requests	2	`429` concurrency_limit

Every accepted request reports where you stand in rate-limit headers.

Chat: POST /v1/chat/completions with system, user, assistant and tool messages. developer is treated as system.
Streaming: stream: true. Add stream_options.include_usage to get token counts in the last chunk.
Tools: tools, tool_choice, parallel_tool_calls.
JSON output: response_format with {"type": "json_object"}.
Sampling: temperature, top_p, stop, seed, presence_penalty, frequency_penalty.
Length: max_tokens or max_completion_tokens, up to 2,048.
Also passed through: logprobs, top_logprobs, top_k, min_p, repeat_penalty.

These return 400 with code unsupported_parameter or unsupported_content:

Thinking mode is off. The model answers directly.

Any other parameter is ignored.