How do I use Gemini 3.5 Flash-Lite in OpenCode or Pi without per-token billing?
In OpenCode or Pi, point at Standard Compute and use google/gemini-3.5-flash-lite: Gemini 3.5 Flash-Lite is on every plan from $19 a month, with EU and US pins, and the plan's compute budget replaces the per-token meter, sending HTTP 402 to new requests at 100%.
| Model | Google's fastest, lowest-cost Gemini 3.5 model (July 2026), described by Google as built for subagent tasks and document parsing. It has a 1,048,576-token input window and 65,536-token max output. |
|---|---|
| Model ID | google/gemini-3.5-flash-liteShort form gemini-3.5-flash-lite also works |
| Base URL | https://api.stdcmpt.com/v1Claude Code uses https://api.stdcmpt.com |
| Model ID, EU pin | google/gemini-3.5-flash-lite:euUS pin: google/gemini-3.5-flash-lite:us, with the same inputs |
| Input | Text, images, files, audio and video |
| Reasoning | No setting; it runs at Google's default |
Copy a working config
OpenCode can keep a stronger main model and hand its small background jobs to Flash-Lite through small_model. Pi and Claude Code pin it for the whole session.
{
"$schema": "https://opencode.ai/config.json",
"model": "standardcompute/standardcompute",
"small_model": "standardcompute/google/gemini-3.5-flash-lite",
"provider": {
"standardcompute": {
"npm": "@ai-sdk/openai-compatible",
"name": "Standard Compute",
"options": {
"baseURL": "https://api.stdcmpt.com/v1",
"apiKey": "{env:STANDARDCOMPUTE_API_KEY}"
},
"models": {
"standardcompute": { "name": "Smart routing" },
"google/gemini-3.5-flash-lite": { "name": "Gemini 3.5 Flash-Lite" }
}
}
}
}
This config reads your key from STANDARDCOMPUTE_API_KEY; export it with a key from the dashboard. Full setup: OpenCode, Pi, Claude Code.
What it costs
| Per 1M tokens | 3.5 Flash-Lite | 3.8 Flash |
|---|---|---|
| Input | $0.30 | $0.75 |
| Cached input | $0.03 | $0.075 |
| Output | $2.50 | $3.75 |
Gemini API paid tier, one rate at every prompt length; Flash-Lite's input rate covers text, image, video and audio. Google lists 3.8 Flash at these rates until December 31, 2026, then $1.50 / $7.50. Source: Gemini API pricing, checked October 13, 2026.
Output costs over eight times input here, so Flash-Lite is most efficient on work that reads a lot and writes little: file triage, search, log reading, document parsing.
What works differently here
- Google's own API has a free tier for Flash-Lite. For light, rate-limited use outside an agent stack, that costs nothing.
- Gemini 3.5 Flash, without Lite, is not served. The served Flash models are Gemini 3.8 Flash and this one.
- Flash-Lite's Artificial Analysis rating is 22: well above the median of 13 for its price tier, and far from a main coding model. Keep it on background and subagent work.
Other ways to run Gemini 3.5 Flash-Lite
| Gemini API$0.30 / $2.50 per 1M | First-party, with a free tier and a batch rate at half price. |
|---|---|
| Vertex AIPay per token | Google Cloud billing, regional endpoints and quotas. |
| Standard ComputeMonthly plan | Flash-Lite with EU and US pins, usable as OpenCode's small_model. |
FAQ
How fast is Gemini 3.5 Flash-Lite?
Artificial Analysis measures about 401 output tokens per second, fourth of 147 models it tracks. A subagent that returns a 2,000-token summary finishes generating in about five seconds.
Can Flash-Lite read audio and video here?
Yes. On Standard Compute it accepts text, images, files, audio and video, the full set Google lists for it.
Sources: Gemini API pricing; Google: Gemini 3.5 Flash-Lite; Artificial Analysis: Gemini 3.5 Flash-Lite
Plans from $19/mo
One key for Gemini 3.5 Flash-Lite and the rest of the catalog, with a monthly compute budget you can watch in the dashboard and no overage charges.