Qwen3.6-35B-A3B on QuickSilver Pro
Qwen 3.6-35B-A3B is Alibaba's MoE workhorse — 35B params, 3B active per token, 262K context window, and strong reasoning at a tiny serving cost. On QuickSilver Pro it is $0.05 input / $0.70 output per million tokens, at parity with the OpenRouter endpoint selected by this route, with no markup or discount.
At a glance
Long-context RAG, document summarization, low-cost MoE with reasoning.
Pricing comparison ($/1M tokens)
| Provider | Input | Output | vs QSP |
|---|---|---|---|
| QuickSilver Pro | $0.05 | $0.70 | — |
| OpenRouter (qwen/qwen3.6-35b-a3b) | $0.05 | $0.70 | same |
| OpenAI (GPT-4o) | $2.50 | $10.00 | 93% lower |
When to use
Qwen 3.6 shines on long-document summarization, multi-document RAG over 100K+ token corpora, and Chinese-language tasks where its training data advantage is real. The 3B-active MoE means it serves cheaply at scale; the 262K context lets you concatenate a Confluence space or technical spec without chunking.
When to use something else
For coding-agent workloads, DeepSeek V4 Flash beats it on HumanEval-style benchmarks; compare input and output rates for your workload. For top-tier reasoning, V4 Pro. QuickSilver Pro serves Qwen 3.6-35B in non-thinking mode: replies are direct, and a `reasoning` field in the request does not turn thinking on.
Quickstart (curl)
curl https://api.quicksilverpro.io/v1/chat/completions \
-H "Authorization: Bearer $QSP_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen3.6-35b",
"messages": [{"role": "user", "content": "Hello!"}]
}'OpenAI-compatible. One-line migration via base_url.
FAQ
It's a 35B-param MoE with only 3B active per token, so it serves cheaply at scale while keeping strong reasoning on math, coding, and long-context retrieval. The 262K context window makes it a natural fit for multi-document RAG and long-document summarization. QuickSilver Pro serves it in non-thinking mode, so replies are direct with no reasoning trace.
No. QuickSilver Pro serves Qwen 3.6-35B in non-thinking mode, so you never pay for a thinking trace. A `reasoning` field in the request does not turn thinking on; for step-by-step reasoning use a model that reasons by default, such as DeepSeek V4 Pro.
$0.05 input / $0.70 output and $0.025 cache read per million tokens, matching the route-selected OpenRouter endpoint (darkbloom fp4) with no markup or discount. The route uses its pinned provider order or the cheapest eligible endpoint under sort: price and its quantization filters; the /models headline can differ. More expensive fallback hosts do not change your price. Use base_url=https://api.quicksilverpro.io/v1 and model="qwen3.6-35b".