Routers charge list price. This is below it.
Model routers earn their margin on top of each provider's list price. Draftworks takes the opposite approach: a small catalog of frontier models, served from Azure at 20 to 25% under what the providers themselves charge, with a signed receipt proving which model answered.
| Per 1M tokens | Provider list | Draftworks Base | Draftworks Scale |
|---|---|---|---|
| GPT-5.6 Solgpt-5.6-sol | |||
| Input | $5.00 | $4.00 | $3.75 |
| Cached input | $0.50 | $0.40 | $0.375 |
| Output | $30.00 | $24.00 | $22.50 |
| GPT-5.6 Terragpt-5.6-terra | |||
| Input | $2.50 | $2.00 | $1.88 |
| Cached input | $0.25 | $0.20 | $0.188 |
| Output | $15.00 | $12.00 | $11.25 |
| GPT-5.6 Lunagpt-5.6-luna | |||
| Input | $1.00 | $0.80 | $0.75 |
| Cached input | $0.10 | $0.08 | $0.075 |
| Output | $6.00 | $4.80 | $4.50 |
| GPT-5.4 Progpt-5.4-pro | |||
| Input | $30.00 | $24.00 | $22.50 |
| Output | $180 | $144 | $135 |
| Grok 4.3grok-4.3 | |||
| Input | $1.25 | $1.00 | $0.938 |
| Cached input | $0.20 | $0.16 | $0.15 |
| Output | $2.50 | $2.00 | $1.88 |
| Kimi K2.7 Codekimi-k2.7-code | |||
| Input | $0.95 | $0.76 | $0.713 |
| Cached input | $0.19 | $0.152 | $0.142 |
| Output | $4.00 | $3.20 | $3.00 |
| Kimi K2.7 Code Fastkimi-k2.7-code-fast | |||
| Input | $1.05 | $0.84 | $0.787 |
| Cached input | $0.18 | $0.144 | $0.135 |
| Output | $4.40 | $3.52 | $3.30 |
List is each provider's published API price, which is what a router bills you before its own fees. Base applies from your first dollar; Scale applies permanently at $5,000 lifetime purchases. Full catalog and rates.
Breadth or unit cost: pick what your workload needs.
Routers are the right tool when you need four hundred models behind one key. The trade is price: aggregation adds a layer that has to be paid for, so you pay list at best.
Draftworks inverts the trade. The catalog is nine models, every one of them served from committed Azure capacity, priced below the provider's own rate, and signed so you can prove nothing was swapped. If your traffic runs on the frontier models, the smaller catalog is simply cheaper.
Common questions.
- How does Draftworks pricing compare to a router like OpenRouter?
- Routers generally pass through each provider's list price and make their margin on top, through platform fees on credit purchases or per-request charges. Draftworks prices below list: every model is 20% under the provider's published price on the Base tier and 25% under on the Scale tier.
- When is a router the better choice?
- When you need breadth. Routers aggregate hundreds of models and handle provider fallbacks. Draftworks serves a deliberately small catalog of frontier models it can price below list and verify cryptographically; if your workload runs on GPT-5.6, Grok 4.3, or Kimi K2.7, the smaller catalog costs meaningfully less.
- How do I know I am getting the model I paid for?
- Every response is signed with Ed25519 over the model id, token counts, and hashes of the exact request and response bytes. Receipts are retrievable forever, and an open audit script compares logprob distributions against the provider's own endpoint. No router offers an equivalent proof.
- Is switching from OpenRouter difficult?
- Both expose the OpenAI-compatible contract, so switching is a base URL and API key change. Model ids map directly: gpt-5.6-sol, grok-4.3, kimi-k2.7-code.
- What does billing look like?
- Prepaid credits from $5, spent at your tier's rates on any model. There are no platform fees on top of the per-token price and no subscription.