Gemini 1.5 Flash-8B

Gemini 1.5 Flash-8B is a chat model from Google and the smallest member of the Gemini 1.5 family. Where Gemini 1.5 Pro sits at the capability end of that generation and Gemini 1.5 Flash trades some quality for speed, Flash-8B goes a step further down the same curve: an 8-billion-parameter distilled model intended for workloads where request volume and response latency matter more than maximum reasoning depth. It is served through Google's Gemini API and Vertex AI rather than as downloadable weights. The model inherits the Gemini 1.5 architecture's long-context design, with a context window measured in the hundreds of thousands to roughly a million tokens depending on the serving endpoint, which is unusual for a model of this size — most sub-10B models cap out far lower. It accepts multimodal input alongside text and returns text output. Because the parameter count is small, per-token generation is fast and the model is inexpensive to run at scale, but it will not match Gemini 1.5 Pro or larger frontier models on multi-step reasoning, hard math, or complex code generation. We do not currently track verified throughput or time-to-first-token figures for this entry; the benchmark fields in our database are unpopulated, so treat any latency expectations as provider-specific. In practice Flash-8B is used as a volume tier: classification, extraction, summarization of long documents, chat over large retrieved contexts, and routing layers that hand harder requests off to a stronger model. Compared with its siblings, the decision is usually between Flash-8B and Flash — Flash-8B is the cheaper and lighter of the two, while Flash offers more headroom on quality. Note that the Gemini 1.5 generation has since been followed by Gemini 2.x models, so check the pricing table on this page for which variants providers still serve.