DeepSeek V4.1 Flash 1017 on QuickSilver Pro
DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash, with 1M-token context. OpenRouter list parity: $0.15 input / $0.60 output per million tokens; cached input is $0.003.
At a glance
Long-context text and images, tools and structured output with optional reasoning.
Pricing comparison ($/1M tokens)
| Provider | Input | Output | vs QSP |
|---|---|---|---|
| QuickSilver Pro | $0.15 | $0.60 | — |
| OpenRouter list price (deepseek/deepseek-v4.1-flash-1017) | $0.15 | $0.60 | same |
When to use
Use this dated snapshot for text and image prompts, automatic tools, forced named functions, strict JSON Schema and streaming with usage. These features were verified on the pinned DeepInfra FP8 endpoint.
When to use something else
The route currently has one qualified provider, so availability depends on that endpoint. Small output budgets can be consumed by reasoning. Audio and video input have not been validated.
Quickstart (curl)
curl https://api.quicksilverpro.io/v1/chat/completions \
-H "Authorization: Bearer $QSP_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4.1-flash-1017",
"messages": [{"role": "user", "content": "Hello!"}]
}'OpenAI-compatible. One-line migration via base_url.
FAQ
No. Use model="deepseek-v4.1-flash-1017" for the October 17 snapshot. The existing deepseek-v4.1-flash route and its prices remain available separately. OpenRouter identifies this snapshot as the model served by DeepSeek under deepseek-flash.
Yes. Reasoning is on by default; pass reasoning: {enabled: false} to disable it, or reasoning: {enabled: true} to enable it. Both forms were tested through the gateway container. Reasoning consumes the completion-token budget; 16- and 64-token probes stayed within the requested total output caps.
$0.15 input / $0.60 output per million tokens, with $0.003 per million cached input tokens. These are the base OpenRouter model-level list rates at factor 1.0. The pinned endpoint has different upstream costs; customer prices use the catalog rates.
DeepInfra standard FP8 is pinned with zero data retention and no provider fallbacks. The endpoint advertises 1M context tokens and 131K output tokens; probes used at most 256. Use base_url=https://api.quicksilverpro.io/v1 with model="deepseek-v4.1-flash-1017" for Chat Completions or Responses.