Model comparisons
Nine head-to-head pages, each built from two real price sets rather than vendor marketing copy: the published list rate and the rate actually billed through this gateway. Where those two disagree about which model is cheaper, the page says so.
Claude vs GPT
Sonnet 5 against GPT-5.6 Terra on blended cost and monthly spend at three volumes.
Claude Opus 5 vs GPT-5.6 Sol
The pair where the list-price ranking flips once gateway rates are applied.
Claude vs Gemini
The cheap tier, where volume amplifies every fraction of a cent.
GPT vs Gemini
The sub-$1 tier compared on billed rates and monthly totals.
Claude Opus 5 vs Sonnet 5
What the 2.5× multiplier costs, and the routing pattern that avoids it.
DeepSeek vs Claude
The widest price gap people actually compare, costed honestly.
Ways to pay for Claude
Direct, prepaid bundle or pay-as-you-go — with the prepay maths.
Lowest-cost LLM APIs
The lowest blended rates in the index, plus every vendor's floor price.
vs OpenRouter
Breadth versus per-token cost, and when running both is the right answer.
How to read these pages
- Two rate columns, one verdict. Official list price and billed rate sit side by side, with the cheaper one marked in each column. A reversal between them is flagged, not buried.
- Monthly cost at three volumes. 10M / 100M / 1B tokens per month, blended at one third output — so you can find the volume where the ranking flips.
- Equal-budget purchasing power. What $100 and $1,000 actually buy in millions of tokens, which is easier to act on than a per-token headline.
- Price structure, not capability. These pages compare cost only. Where a capability difference matters, they point at the vendor's own documentation rather than guessing.
If your pair is not here
The cost calculator covers every model in the index and blends them at your own input/output ratio. Set your ratio first — output tokens cost several times more than input, so a comparison at the wrong ratio can invert the answer. The three-mix ranking shows how far that inversion can go.
FAQ
Which comparison should I start with?
If you have already chosen one model, open the page for it: Claude vs GPT covers the most common shortlist, and Claude vs Gemini the sub-$1 tier where volume matters most. If you have not chosen yet, start from the lowest-cost list and work down.
Do these pages use official list prices or your own rates?
Both, listed separately on the same table. Where a model has a published vendor list rate we show it next to the rate we bill, and when the two rankings disagree we say so explicitly — that reversal is usually the most useful line on the page.
Why does the winner change between the two rate sets?
Because a list rate and a billed rate move on different calendars. Vendor list prices change when the vendor decides; billed rates follow upstream promotions. 16 models have a page each if you want the full picture rather than a head-to-head.
Why do only some models show a list price?
16 models have a detail page with the rate we bill. Only a subset of those carry a published list rate we can cite, and only those can be cross-checked — the rest are shown at face value with no comparison implied. We do not estimate a list price that nobody publishes.
Can I compare two models that are not listed here?
Yes — the cost calculator accepts every model in the index and blends them at your own input/output ratio, which is a fairer comparison than a per-token headline. If you want a head-to-head written up, the pairs above are the ones people search for most.
Does the cheaper model always cost less in practice?
No. Output pricing carries most of a real bill, so a model that wins on input rate can lose on a workload that generates a lot of text. Claude Fable 5 is the most expensive output rate in the catalogue at $25 per 1M, while GPT-5.6 Luna starts at $0.1 per 1M input — a gap that only matters if your workload looks like one or the other.