Google: Gemini 2.5 Flash Lite

At $0.10 per 1M input tokens, this lightweight reasoning model in the Gemini 2.5 family is cheaper than roughly 75 percent of comparable models, and it is built for throughput and cost efficiency rather than peak capability. It takes a prompt, reasons briefly if needed and returns an answer at very low latency and cost, handling images and structured output along the way. A support team auto-summarising a large ticket backlog, or a pipeline classifying documents at scale, can run continuously without the bill climbing, while a 1,048,576-token context window still lets each request carry long context. Vision input, chain-of-thought reasoning and structured JSON output make it usable in real automation, and a 2025-01-31 knowledge cutoff marks its limit. Among its alternatives the positioning is the stable, low-cost workhorse: unlike Gemini 3.5 Flash, which pays more for near-Pro coding and reasoning, this favours speed and price, and unlike the dated Flash Lite Preview it names the general release rather than a snapshot. The one caveat is the lack of a free tier, usage is billed per token, though at a rate low enough that trialling costs little. Ideal for content-creation and research teams needing dependable, low-cost inference at volume. Choose it when cost and latency lead; move up to a stronger Flash or Pro tier when a task demands more reasoning than a lite model can give.