Normalized public GPT/Codex and Claude relay multipliers, route-level distributions, subscription utilization, and pricing implications.
· 10 min read
Loading article…
Find models, providers, pricing pages, and practical guides.
Competitor pages use incompatible units: an internal 0.2x group, a direct “95% off,” a one-yuan platform dollar, or a daily-reset subscription. This analysis normalizes each public price to the customer’s CNY spend divided by the official API price converted at US$1 = ¥7.
Across four publicly reproducible pay-as-you-go competitors, seven GPT/Codex routes have a median of 0.071×; twelve Claude routes have a median of 0.189×. The median Claude offer is about 2.65 times the GPT median.
The core sample includes logged-out public pricing that is model-family specific and convertible to an official-CNY multiplier. For platforms where ¥1 buys one platform dollar and the internal group rate is r, the normalized rate is r ÷ 7. Direct official-versus-sale price cards use sale ÷ official. Subscription full-use values are analyzed separately because utilization changes the realized rate.
| Metric | GPT / Codex | Claude |
|---|---|---|
| Comparable public routes | 7 | 12 |
| Minimum | 0.029× | 0.086× |
| P25 | 0.054× | 0.100× |
| Median | 0.071× | 0.189× |
| P75 | 0.079× | 0.264× |
| Maximum | 0.086× | 0.286× |
GPT is tightly compressed; 86% of routes fall between 0.05× and 0.10×. Claude is split between lower-cost third-party routes and higher-cost Max/client-compatible routes. Giving each provider one vote produces roughly 0.064× for GPT and 0.189× for Claude, so multi-group providers do not change the conclusion.
| Provider | Public route | Effective multiplier |
|---|---|---|
| API Fast Link | Codex Pro / Codex-only / external | 0.029× / 0.071× / 0.086× |
| Pateway | GPT-5.5 | 0.050× |
| Unity2 | Codex | 0.057× |
| APIKEY.FUN | Codex-only / external | 0.071× / 0.086× |
GPT / Codex public PAYG ranges
Each provider's public groups normalized to official CNY API pricing.
GPT / Codex distribution
Seven comparable public PAYG routes.
| Provider | Public route range | Effective multiplier |
|---|---|---|
| API Fast Link | Max, Lite, CC-only, external | 0.086×–0.286× |
| APIKEY.FUN | Lite, Plus, CC-only, external | 0.100×–0.286× |
| Unity2 | Kiro, CC Max, Max | 0.086×–0.286× |
| Pateway | Sonnet / Opus | 0.150× |
Claude public PAYG ranges
Low-cost third-party groups and Max/client-compatible groups form two visible bands.
Claude distribution
Twelve comparable public PAYG routes.
| Provider | Product | GPT full-use rate | Claude full-use rate | Main caveat |
|---|---|---|---|---|
| Unity2 | daily shared quota | about 0.017× | about 0.034× | derived; quota must be fully used |
| Codesome | 30-day daily quota | — | 0.044×–0.046× | near-100% utilization |
| LetAiCode | expiring credit packs | 0.042×–0.071× | 0.252×–0.428× | expiry and dynamic exchange rates |
| AIGoCode | four-week allowance | 0.121×–0.130×* | 0.121×–0.130×* | assumes platform dollars map to official API value |
Codesome's 0.044×–0.046× headline becomes 0.087×–0.093× at 50% utilization and 0.175×–0.185× at 25%. A subscription's theoretical floor is therefore not the typical customer's realized rate.
The account-pool sample is only the low-cost segment. AIGoCode's general PAYG anchor is about 0.143×; OpenModel publishes about 0.55× for OpenAI and 0.95× for Anthropic; Unity2's official-key group normalizes to about 1.071×. The market contains account-pool, aggregator, and official-key products with different reliability and compliance boundaries.
With a 10% direct-loss reserve:
| Supply cost | Reference retail | Contribution margin | Market position |
|---|---|---|---|
GPT 0.05× | 0.086× | about 32% | top of comparable market |
GPT 0.05× | 0.10× | 40% | about 17% above sample maximum |
GPT 0.10× | 0.20× | 40% | 2.3× the sample maximum |
Claude 0.08× | median 0.189× | about 48% | commercially comfortable |
Claude 0.12× | median 0.189× | about 27% | needs roughly 0.24× for 40% |
Claude 0.15× | P75 0.264× | about 33% | premium band |
GPT cost at 0.05× is workable but cannot simultaneously match the market median and preserve 40% contribution margin. GPT cost at 0.10× should not compete as a generic low-cost route. Claude at 0.08×–0.10× fits an economy product around 0.16×–0.20×; Claude at 0.12×–0.15× needs a stable/external-access tier around 0.24×–0.30×.
The core distribution includes API Fast Link, APIKEY.FUN, Unity2, and Pateway. Sponsor-copy-only or login-only values are excluded from the statistics. Identical multipliers can still hide differences in client restrictions, cache billing, model fidelity, context, rate limits, refunds, account survival, and SLA.
Prices were checked on July 11, 2026. Primary public sources: API Fast Link, APIKEY.FUN groups API, Unity2 groups API, Pateway homepage API, LetAiCode packages, LetAiCode model rates, Codesome, and AIGoCode.