Provider APIGenerationRecommendedVerified 2026-11-05
Gemini 3.5 Flash-Lite
gemini-3.5-flash-lite
Gemini 3.5 Flash-Lite is a language model. Called on Mixpeek through the provider API.
Mixpeek calls the provider's API by name. Provider terms apply.
New opens the Designer in Studio. For an existing namespace, open it in Studio and use its Designer tab.
No demo for Gemini 3.5 Flash-Lite yet
A video, image, slideshow or playground goes here once we record one on our own data.
What it is good and bad at
- Best at
- The platform's default LLM (DEFAULT_GEMINI_MODEL) for cheap per-document work in retriever stages (enrich, filter, summarize) and for video-segment description and OCR in multimodal_extractor v1.
- Worst at
- Multi-step reasoning and long structured extraction, where the code picks gemini-2.5-flash or escalates to gemini-2.5-pro; per-document cost grows with candidate count in llm_filter and llm_enrich.
- Task
- Generation
- Input
- Text, Image, Video, Document
- Output
- Text, Structured data
- Use cases
- Video understanding, Tagging, Captioning, OCR, Brand detection
- Why we recommend it
- Platform default Gemini model since 2026-10-10 (shared/models/llm/gemini_models.py).
Templates that use it
No template uses this model yet. Browse templates.
Input and output
Input
- text
- prompt plus document fields
- image
- URLs or inline data
- video
- segments (multimodal_extractor v1)
Output
- text
- free text, or JSON when response_shape / response_schema is set
How Mixpeek wires it
- Extractors
multimodal_extractor@v1fixed (video description and OCR when run_video_description or run_ocr)universal_extractor@v1fixed (GENERATION_MODEL)scrolling_text_extractor@v1fixed (direct Gemini OCR fallback, _OCR_FALLBACK_MODEL)
- Feature URIs
- None today
- Vector indexes
- None
- Retriever stages
- llm_enrich, llm_filter, summarize, agent_search, agentic_enrich, rerank
- Combines with
- feature_search -> llm_filter or llm_enrich (default model_name) -> summarize
- rerank with inference_name mixpeek__decision and decision_model=gemini-3.5-flash-lite
How it runs
- Status
- Provider API Provider API. Retriever stages and extractors call it through CentralLLMService or google-genai directly; universal__api routes any gemini-* model for POST /v1/inference.
- Serve apps
universal__api- Runtime
- Gemini developer API or Vertex AI (clients.providers.google.gemini_backend, shared.models.llm GoogleProvider)
- Weights
- Provider API, no remote code
- Parameters
- Not recorded
- Size on disk
- Not recorded
- Preprocessing
- Prompt templates per stage; media passed as URLs or inline data.
- Postprocessing
- Text, or JSON parsed against the requested schema.
- Serving
- both
- Hardware
- Provider API
Scale to zero
- Minimum replicas
- 0
- Maximum replicas
- 6
- Idle timeout
- 600 s
- Memory
- Not recorded
- CPUs
- 0.1
- GPUs
- 0
- Cold start, node warm
- Not recorded
- Cold start, new node
- Not recorded
Measured cost
Unmeasured
Billed per token by Google; no committed throughput and cost measurement.
Our evaluations
Unmeasured
We have not evaluated this model on our own data yet. Publisher results are in the sources below.
Questions
- What is Gemini 3.5 Flash-Lite best at?
- The platform's default LLM (DEFAULT_GEMINI_MODEL) for cheap per-document work in retriever stages (enrich, filter, summarize) and for video-segment description and OCR in multimodal_extractor v1.
- What are the limits of Gemini 3.5 Flash-Lite?
- Multi-step reasoning and long structured extraction, where the code picks gemini-2.5-flash or escalates to gemini-2.5-pro; per-document cost grows with candidate count in llm_filter and llm_enrich.
- Can Gemini 3.5 Flash-Lite be used commercially?
- Yes. Paid Google API under Google's terms; tenants can bring their own key.
- How does Gemini 3.5 Flash-Lite run on Mixpeek?
- Provider API. Retriever stages and extractors call it through CentralLLMService or google-genai directly; universal__api routes any gemini-* model for POST /v1/inference.