AI Gateway Models

Reference for all models available through Salad AI Gateway, including context windows, capabilities, and recommended use cases.

Last Updated: September 2, 2026

Salad AI Gateway is currently in beta.

Available Models

qwen3.6-35b-a3b

Property	Value
Model ID	qwen3.6-35b-a3b
Model name	Qwen3.6 35B-A3B
Context window	262,144 tokens
Max input	262,144 tokens
Max output	262,144 tokens
Input	Text, image
Output	Text

Best for: Agentic tasks, complex multi-step reasoning, code generation, instruction following.

This is the recommended model for most agentic and coding use cases. It handles tool calling reliably, supports structured output, and performs well in long agentic sessions.

---

qwen3.5-35b-a3b

Property	Value
Model ID	qwen3.5-35b-a3b
Model name	Qwen3.5 35B-A3B
Architecture	Mixture of Experts (MoE)
Active params	~3.5B (35B total)
Context window	131,072 tokens
Max output	16,384 tokens
Input	Text, image
Output	Text

Best for: Agentic tasks, complex multi-step reasoning, code generation, instruction following.

The Mixture of Experts architecture activates only a subset of parameters per token, providing strong reasoning capability at lower latency than a dense 35B model. It handles tool calling reliably and performs well in long agentic sessions.

---

Choosing a Model

Use Case	Recommended Model
Agentic coding (Cline, Roo Code)	qwen3.6-35b-a3b
Complex reasoning / planning	qwen3.6-35b-a3b
Cost-efficient agentic tasks	qwen3.5-35b-a3b

API Usage

Specify the model ID in the model field of your request:

curl https://ai.salad.cloud/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $SALAD_AI_GATEWAY_API_KEY" \
  -d '{
    "model": "qwen3.6-35b-a3b",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

To list available models programmatically:

curl https://ai.salad.cloud/v1/models \
  -H "Authorization: Bearer $SALAD_AI_GATEWAY_API_KEY"

Notes

• All models support the /v1/chat/completions endpoint with streaming ("stream": true).
• All models support text and image inputs, text outputs, tool calling, structured outputs, and reasoning.
• Context window limits are enforced server-side. If your input exceeds the limit, the request will return an error.
• The model catalog may expand during beta.