Ling-3.1-flash API pricing and cost calculator

How this is calculated

Because the model suits volume, the calculator opens on a smaller request than most pages: 20,000 input tokens, 1,000 output tokens, half the input cached, and 10,000 requests a month. One uncached call costs $0.0069. With half the input read from cache it's $0.0045, and the starting total for 10,000 calls is $45. While the deal lasts, the same half cached call costs $0.001125. The deal is a promotion, not a permanent rate, so keep the standard $0.0045 figure in any plan that has to hold after it ends. Now try workloads that match a cheap text model. A support desk that triages 500,000 tickets a month, with 1,500 input tokens and 100 output tokens per ticket and no caching, costs $270 a month. A code review bot that reads 30,000 tokens per pull request with a shared prefix cached for half the input, and writes 1,200 tokens of comments, costs $129.60 for 20,000 pull requests. A one off summary of a very long report, 400,000 input tokens and 2,000 output tokens, costs $0.1218 uncached. Since the model takes text input only on the API listing, anything with images or audio needs a transcription or captioning step first, and that step has its own cost. Build it into the estimate. The weights are published under the MIT license in two versions, Ling-3.1-flash and an upgraded Ling-3.1-flash-1019. Which one a host serves can affect results, so confirm the checkpoint before you compare quality across providers. The prices shown are the maker's own. Open weights allow other hosts, and the calculator will show a lower host price here when one is listed.

Verdict

Ling-3.1-flash fits classification, triage, extraction and other text work where you send many calls and each answer is short. Pricing is only half of it. Run a sample of your own traffic, count how many answers pass your checks, and divide spend by passes. A cheap call that needs a retry doubles its own cost. Keep deal prices out of long term plans, check the host and checkpoint you're really using, and read the Hugging Face model card for the details of the MIT licensed weights. If you'd rather run it yourself, the VRAM calculator page for this model shows the memory budget.