DeepInfra and Together listed prices
These are different models. The table is here so the two price lists can be read side by side. It is not a claim that a DeepInfra row replaces a Together row.
| Host | Id as printed | Input / 1M | Output / 1M | Context label |
|---|---|---|---|---|
| DeepInfra | DeepSeek-V4-Flash-0731 | $0.06 (cached input $0.015) | $0.18 | 1024k |
| DeepInfra | phi-4 | $0.07 | $0.14 | 16k |
| Together | zai-org/GLM-5.3-Flash | $0.15 (cached $0.03) | $0.50 | 1M |
| Together | moonshotai/Kimi-K3 | $3.00 (cached $0.30) | $15.00 | 1M |
DeepInfra writes context as 1024k and 16k on the pricing table. Together writes 1M on the model page. This page does not convert those labels into a shared token integer. DeepInfra’s row links are /deepseek-ai/DeepSeek-V4-Flash-0731 and /microsoft/phi-4. The pricing page does not, in the cells copied here, say either DeepInfra model is for coding. Together’s GLM page uses the heading “Frontier Coding at Flash Cost”.
Directory: DeepInfra and Together. Token arithmetic for these rates is on the cost calculator.
Other host notes
- Groq and Cerebras, from their docs
- OpenRouter and Together, from their docs
- Token price calculator
- Context windows
- LLM APIs category list
Sources
Last checked 2026-10-22.