Serverless Inference Invoice from Digital Ocean (2026-07)
Total Usage Charges: $0.51 from Digital Ocean Serverless Inference
I recently experimented with several large language models through a serverless inference platform, including DeepSeek R1 Distill Llama 70B, Llama 3.3 70B, Qwen 3.5, Qwen3-32B, Kimi K2.5/K2.6 and MiniMax M2.5. Despite trying multiple models and processing approximately 382,000 tokens, the total bill came to only $0.51. This shows just how affordable serverless AI inference can be when there is no need to reserve or maintain dedicated GPU infrastructure.
Most of the expense came from output tokens rather than input tokens. The largest individual charge was $0.20 for generating 100,302 tokens with Kimi K2.5. By comparison, MiniMax M2.5 produced a similar 100,933 output tokens for only $0.09, mainly because its off-peak output rate was considerably lower. Input-token charges were almost negligible, with most individual items rounded to just one cent.
The invoice also highlights how much model pricing can vary, even when the workloads are broadly similar. Choosing a model is therefore not only about intelligence, speed and output quality; token pricing and off-peak discounts can make a substantial difference at scale. For small experiments, the difference may only be a few cents, but for applications generating millions or billions of tokens, selecting the right model and running workloads during off-peak periods could lead to significant savings.
Steem to the Moon🚀!
- You can rent Steem Power via rentsp!
- You can swap the TRON:TRX/USDT/USDD to STEEM via tron2steem!
- You can swap the STEEM/SBD to SUI via steem2sui!
- You can swap the STEEM/SBD to SOL Solana via steem2sol!
- You can swap the STEEM/SBD to ETH Ethereum via steem2eth!
- You can swap the STEEM/SBD to Tether USDT (TRC-20) via steem2usdt!
- You can swap the STEEM/SBD to TRX (TRON) via steem2trx!
- You can swap the STEEM/SBD to BTS (BitShares) via steem2bts!
- Register a free STEEM account at SteemYY!
- Steem Block Explorer
- ChatGPT/Steem Integration: You can type !ask command to invoke ChatGPT
- Steem Witness Table and API
- Other Steem Tools
Support me, thank you!
Why you should vote me? My contributions
Please vote me as a witness or set me as a proxy via https://steemitwallet.com/~witnesses
Your experiment highlights the flexibility and cost-effectiveness of serverless inference, especially when combined with the right model. I'm curious, how do you plan to apply these findings to future AI model testing? 💻🤖