Gemini 2.5 Pro on QuickSilver Pro
Gemini 2.5 Pro is Google's proven pro-tier model — 1M-token context, multimodal, thinks deeply on hard problems. On QuickSilver Pro it lists at $1.0625 input / $8.50 output per 1M tokens, ~15% below Vertex retail ($1.25/$10). A cheaper, GA-stable step below the 3.x Pro previews when you want strong reasoning without preview-tier risk or price.
At a glance
Proven pro-tier reasoning at GA stability — deep thinking, 1M context, well under the 3.x Pro previews on price.
Pricing comparison ($/1M tokens)
| Provider | Input | Output | vs QSP |
|---|---|---|---|
| QuickSilver Pro | $1.06 | $8.50 | cheapest |
| OpenRouter (google/gemini-2.5-pro) | $1.25 | $10.00 | 15% cheaper |
| OpenAI (GPT-4o) | $2.50 | $10.00 | 15% cheaper |
When to use
Reach for 2.5 Pro when you need solid pro-tier reasoning on a long-stable model: complex analysis, multi-step agents, code review and refactors, and long-context document work where 2.5 Flash isn't quite strong enough. It's the safe, cheaper reasoning workhorse — GA and long-stable, predictable behavior, and well below the 3.1 Pro preview on price.
When to use something else
For routine high-volume chat, 2.5 Flash ($0.255/$2.125) or 3.1 Flash Lite ($0.2125/$1.275) are far cheaper. For the very hardest reasoning or newest capabilities, 3.1 Pro Preview ($1.70/$10.20) or 3.5 Flash ($1.275/$7.65) push further. 2.5 Pro thinks by default and `reasoning.enabled=false` is silently dropped — for predictable, non-thinking token budgets use 2.5 Flash Lite.
Quickstart (curl)
curl https://api.quicksilverpro.io/v1/chat/completions \
-H "Authorization: Bearer $QSP_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gemini-2.5-pro",
"messages": [{"role": "user", "content": "Hello!"}]
}'OpenAI-compatible. Same model as OpenRouter; one-line migration via base_url.
FAQ
3.1 Pro Preview is the newer flagship — deeper reasoning on the hardest tasks, $2/$12 per 1M at Vertex retail, but preview (behavior and pricing can shift). 2.5 Pro is the GA-stable pro tier at a notably lower cost ($1.25/$10 retail vs $2/$12). Use 2.5 Pro for production reasoning that values stability and price; escalate to 3.1 Pro for the hardest evals.
Yes — 2.5 Pro thinks by default and produces a reasoning trace before its answer, which you pay for in output tokens. The `reasoning.enabled=false` flag is silently dropped through QuickSilver Pro (Vertex doesn't surface the thinking toggle). Budget output tokens accordingly, or use 2.5 Flash Lite for a genuinely non-thinking Gemini.
QuickSilver Pro lists 2.5 Pro at $1.0625 input / $8.50 output per 1M tokens — ~15% below Vertex retail's $1.25/$10. One OpenAI-compatible key across 18 models, one bill, referral rewards, and a `usage.cost` field on every response for per-request accounting.