gemini-3.5-flash-lite) is Google’s lightweight multimodal model updated in July 2026 (stable release), built for high-frequency, low-latency, low-cost workloads. It accepts text/image/video/audio/PDF input with a 1M context window and 64K output. APIYI has completed a full dual-endpoint test pass (25+5 cases): both the native Gemini format and the OpenAI-compatible format work out of the box, and native tools — Search grounding, URL context — are verified working.
Available on APIYI now: model name
gemini-3.5-flash-lite, in the default / svip groups. Unlike 3.6 Flash — no thinking output by default, simple requests return in about 2 seconds in our tests; pass thinkingLevel: "high" to enable deep reasoning explicitly.Highlights
Best-in-class value
$0.30 input / $2.50 output per 1M tokens (audio input at the same rate) — a fifth to a third of 3.6 Flash. Built for high-frequency and batch workloads.
Zero thinking, minimal latency
No thinking tokens by default; simple requests measured at ~2s (vs ~4.5s on 3.6 Flash). Ready out of the box for support bots, classification, and extraction.
Full multimodal understanding
Image, PDF, and audio verified accurate (video shares the same pipeline), with the same 1M context as the flagship.
Native tools available
Google Search grounding, Maps grounding, URL context, and code execution verified working on the native endpoint — no Google API Key required.
Model details
Verified capability matrix
APIYI test results from July 22, 2026 (official claims vs. measured behavior):Pricing
Pricing note: APIYI matches official pricing; the discount comes from top-up bonuses: +10% on $100, up to +20% (≈17% off). See top-up promotions.
Controlling thinking
Opposite of 3.6 Flash — this model does not think by default, which is exactly why it’s fast and cheap. Measured:Quick start
Native Gemini format (recommended — full tool support)
OpenAI-compatible format (drop-in for existing code)
FAQ
When should I pick Flash-Lite over 3.6 Flash?
When should I pick Flash-Lite over 3.6 Flash?
Pick Flash-Lite for throughput-first workloads — high-frequency Q&A, classification, extraction, translation, support bots (about twice as fast, down to a fifth of the cost). Pick 3.6 Flash for deep reasoning, complex planning, or Computer Use.
Do I need a Google API Key for the native endpoint?
Do I need a Google API Key for the native endpoint?
No. Put your APIYI token (the
sk- key) in the x-goog-api-key header. With the official google-genai SDK, just set base_url to https://api.apiyi.com.How do I monitor thinking spend?
How do I monitor thinking spend?
On the native endpoint, read
usageMetadata.thoughtsTokenCount. Note the OpenAI-compatible endpoint does not report reasoning_tokens for this model — use the native endpoint for precise observation.Does implicit caching hit?
Does implicit caching hit?
We observed zero hits over three back-to-back requests with a 15.9K-token shared prefix (3.6 Flash hit once under the same conditions). Don’t build cost models on cache hits. The explicit cache API is not yet enabled on the platform.