Google: Lyria 3 Pro Preview

Google: Lyria 3 Pro Preview is a text-and-image input model from Google, currently available at no cost. It supports a context window of 1,048,576 tokens and can generate up to 65,536 tokens per response. It does not support tool use, reasoning modes, or structured output, which limits its utility for agentic pipelines or applications that require predictable, schema-bound responses. Because it carries a price of zero, it is worth shortlisting for experimentation, prototyping, or workloads where budget is the primary constraint. The tradeoff is transparency: there is no independent benchmark coverage to consult, so performance relative to competing models is genuinely unknown. Teams with strict quality thresholds or production reliability requirements should treat it as unproven until third-party evaluations emerge, while cost-sensitive users exploring multimodal text generation have little to lose by testing it.