LLM cost calculator
Price any workload on any of the catalogue models, then see what caching, exact repeats and cascade routing take off.
On the default scenario, gemini-3.1-pro costs $760 a month at list and $323 through TokenZero with the prefix cached, 10% exact repeats and 40% of prompts cascaded, a 57.55% saving; gpt-5.6-luna goes from $76 to $44 (41.97%). Change any input and the table recomputes from pricing table 2026-10-26.
· Source: pricing table 2026-10-26 and the gateway's own tokenizer and benchmark files; your inputs never leave the browser except token counts, which are not stored
At a glance
- Models available
- 62
- Model A · B
- gemini-3.1-pro · gpt-5.6-luna
- gemini-3.1-pro at list / through TokenZero
- $760 / $323 per month (computed)
- gpt-5.6-luna at list / through TokenZero
- $76 / $44 per month (computed)
- Saving on the default scenario
- 57.55% · 41.97% (computed)
- Cascade siblings
- gemini-3.8-flash · —
- Engines modelled
- prompt-cache formatting, exact cache, cascade routing
- Not modelled
- semantic cache hits, long-context tiers, processing-tier multipliers
- Parity floor
- 99.8%
Default scenario: 2,000 input (1,500 static) / 300 output tokens, 100,000 calls per month, 10% exact repeats, 40% simple prompts
| Model | Per call (list) | Per month (list) | Prefix cached | + exact repeats | + cascade | Saving |
|---|---|---|---|---|---|---|
| gemini-3.1-pro | $0.00760 | $760 | $490 | $441 | $323 (via gemini-3.8-flash) | 57.55% |
| gpt-5.6-luna | $0.00076 | $76 | $49 | $44 | $44 | 41.97% |
Calculator
Questions people ask
How is the TokenZero estimate computed?
List cost per call from the pricing table; the static prefix priced at the cached-input rate when it clears the model's minimum; exact repeats at $0; the cascaded share priced on the cheaper sibling. Semantic hits and processing tiers are left out, so the estimate is conservative.
Why can the saving differ from my receipts?
Receipts use the provider-reported usage: real cache hit rates, real repeat share and the classifier's real simple-prompt share for your traffic. The calculator is arithmetic on your assumptions.
Which models can I compare?
Any two of the 62 models in pricing table 2026-10-26: OpenAI, Anthropic, Gemini, xAI and DeepSeek.
What does TokenZero cost?
Free up to $500 per month of routed spend; then 15% of verified savings with a 3x-ROI-or-$0 rule. Gateway overhead is 1.52 ms p50, measured.