Gemini 3 Flash Preview on QuickSilver Pro
Gemini 3 Flash Preview is the next-gen Flash from Google — meaningfully stronger reasoning than 2.5 Flash while staying in the Flash price tier. 1M-token context, thinks on hard problems. On QuickSilver Pro it's $0.425 input / $2.55 output per 1M tokens, ~15% below Vertex retail and OpenRouter's $0.50/$3.00. Preview API; treat output formats as in-flux until Google promotes to GA.
At a glance
Stronger-reasoning Flash at the Flash price tier — for teams who want 3.x quality without 3.x Pro pricing.
Pricing comparison ($/1M tokens)
| Provider | Input | Output | vs QSP |
|---|---|---|---|
| QuickSilver Pro | $0.42 | $2.55 | cheapest |
| OpenRouter (google/gemini-3-flash-preview) | $0.50 | $3.00 | 15% cheaper |
| OpenAI (GPT-4o-mini) | $0.15 | $0.60 | 325% more expensive |
When to use
Pick 3 Flash Preview when 2.5 Flash isn't smart enough but 3.1 Pro is overkill: harder coding turns in an agentic loop, multi-step analysis where Flash 2.5 fails some evals, long-context summarization with non-trivial cross-document reasoning. The price is ~67% higher than 2.5 Flash but still ~4× cheaper than 3.1 Pro Preview on output.
When to use something else
For revenue-critical production paths, prefer a Gemini GA model (2.5 Flash) until 3 Flash is promoted — preview semantics can change. For cheap high-volume chat, 2.5 Flash Lite at $0.085/$0.34 is ~6× cheaper. For top-tier reasoning, escalate to 3.1 Pro Preview ($1.70/$10.20) or DeepSeek R1 ($0.56/$2.00).
Quickstart (curl)
curl https://api.quicksilverpro.io/v1/chat/completions \
-H "Authorization: Bearer $QSP_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gemini-3-flash-preview",
"messages": [{"role": "user", "content": "Hello!"}]
}'OpenAI-compatible. Same model as OpenRouter; one-line migration via base_url.
FAQ
Preview means Google can change output formats, thinking behavior, or pricing without prior notice. For prototyping and A/B evals against 2.5 Flash, ship today. For paths where a behavior shift would break customers, gate behind a feature flag and pin to 2.5 Flash GA as the fallback until Google labels 3 Flash as GA.
Google's positioning is 'stronger reasoning at the same Flash latency tier' — 3 Flash thinks more aggressively by default and lands better on coding/reasoning evals than 2.5 Flash. The price is ~67% higher ($0.425/$2.55 vs $0.255/$2.125 per 1M). Whether the quality bump justifies the cost is a per-workload question — run a side-by-side eval on your traffic before committing.
QuickSilver Pro lists Gemini 3 Flash Preview at $0.425 input / $2.55 output per 1M tokens — ~15% below Vertex retail and OpenRouter's $0.50/$3.00. OpenAI-compatible API; switch from another provider with a base_url + key swap on the OpenAI SDK.