Claude API relay services 81 in total

Claude draws the most concentrated demand of any family from developers here, and the reason is not chat — it is Claude Code. That command-line tool turned "a few million tokens a day" into a normal workload, which makes unit price and stability unusually sensitive.

There is one technical dividing line you cannot avoid when choosing a Claude relay: whether it supports Anthropic's native protocol. Claude Code requests /v1/messages, not OpenAI's /v1/chat/completions. A service that only implements the OpenAI protocol chats fine in a browser and returns 404 the moment Claude Code connects. So the test is simply: send one request to /v1/messages.

The second dividing line is whether prompt caching is passed through properly. Claude Code re-sends the same context constantly, and the official price for a cache hit is roughly a tenth of a normal input token. If the relay strips the cache fields in the middle, your bill can come out several times higher while the headline rate stays identical — something you will not spot on a marketing page.

One more caution: some services advertise "the full Claude line" while Opus is not actually usable, usually because it costs them more — it gets quietly downgraded to Sonnet or returns an error outright. Our API probe records what /v1/models actually returns, which is a reasonable starting point for cross-checking.

• Most stable：Highest current uptime. Where samples are thin we show the public-dataset baseline instead, and the detail page states the source.
• Fastest：Home-page response time from the most recent probe. A single measurement is noisy, so treat it as a range rather than an absolute.
• Broadest model coverage：This uses the model count our endpoint probe actually detected, not the support list the service advertises — the two are measured differently, and the advertised figure is usually the more optimistic one.

I configured a relay but Claude Code still returns 404. Why?

Most likely the service does not implement Anthropic's native protocol. Claude Code uses /v1/messages, not the OpenAI-style /v1/chat/completions. To check: point base_url at the service and send a single request to /v1/messages — a 404 or 405 means it is not implemented.

A Claude service with a cheap rate — why is the bill higher?

The most common reason is that prompt caching is not passed through. Cache hits in Claude Code save roughly nine tenths of the repeated input cost; without support you pay full price for every re-sent context. The rate looks low on the surface, and the total can still be higher than a pricier service that supports caching.

How do I tell whether the Opus I am getting is real?

Cross-check in three directions: request /v1/models and look for Opus in the list; ask a hard question that only a flagship answers correctly and compare with the official channel; and watch latency — Opus responds noticeably slower than Haiku, so a model labelled Opus that answers at Haiku speed is worth suspecting.

Running Claude Code on it — should I worry about concurrency limits?

Yes. Claude Code fires off several sub-task requests in parallel, so instantaneous concurrency is far higher than in browser chat. Some services set a low per-account ceiling, which shows up as a burst of mid-task failures rather than a clean rate-limit error. Run one complete medium-sized project first and watch the error rate.