Claude Fable 5 and GPT-5.6 Sol pricing on Relay365

Splitting a first top-up across two duty-desk lanes

From the Relay365 duty desk, the simplest way to start is arithmetic rather than guesswork. Relay365 is an independent third-party relay that puts every lane on one table: Claude, GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax and Hunyuan traffic all run through one prepaid wallet, one set of API keys and one usage log. You top up, you pick a lane, and the desk keeps the meter honest.

For this walkthrough we split a first top-up between two lanes only: Claude Fable 5 (wire name claude-fable-5) and GPT-5.6 Sol (wire name gpt-5.6-sol). Both are priced per 1 million tokens in approximate USD, while the account itself is kept in CNY. Four token types are priced separately on every lane: input, output, cache write and cache read. That separation is the whole point of the split, because the two models charge very differently for the same conversation.

Claude Fable 5 costs $3.79 per 1 million input tokens, $18.94 per 1 million output tokens, $4.66 for a cache write and $0.2917 for a cache read. GPT-5.6 Sol costs $0.3736 input, $2.09 output and $0.0292 cache read per 1 million tokens. A new account also receives $0.25 in signup credit, small but real, and the wallet carries no monthly fee. Every call is settled on the token counts the upstream reports.

Two billing details matter when you are watching a first balance. A call that fails upstream with a 502 or a 503 is not billed at all. A disconnected stream is billed only for what was actually used. The desk also enforces clear traffic limits: more than 120 calls in 60 seconds on one key, or more than 8 requests in flight for one account, returns HTTP 429.

What $50 on each lane buys, token by token

Here is the plain-numbers version. For example, if you split a $100 first top-up evenly, $50 goes to the Claude Fable 5 lane and $50 to the GPT-5.6 Sol lane. At Claude Fable 5 rates, $50 covers about 13.19 million input tokens or about 2.64 million output tokens. At GPT-5.6 Sol rates, the same $50 covers about 133.83 million input tokens or about 23.92 million output tokens. Nothing is estimated behind the scenes; these are the published per-1-million-token rates applied to your own balance.

Mix the two token types and the difference stays visible. For example, if you send 1 million input tokens and 1 million output tokens, Claude Fable 5 costs $22.73 ($3.79 + $18.94) while GPT-5.6 Sol costs $2.4636 ($0.3736 + $2.09). Against official list prices — $10 input and $50 output for Claude Fable 5, $4 input and $20 output for GPT-5.6 Sol — the same two calls would be $60 and $24. The Relay365 output rate for Claude Fable 5 is about 38% of official; for GPT-5.6 Sol it is about 10% of official.

Caching sharpens the split further. A Claude Fable 5 cache write is $4.66 per 1 million tokens while a cache read is $0.2917, so a workflow that repeats context reads far cheaper than it writes. GPT-5.6 Sol cache reads land at $0.0292 per 1 million tokens. Because Relay365 prices all four token types separately and reports them separately, you can see exactly where a month of usage went instead of receiving one blended line.

The ledger also keeps the exit open. Within 14 days of a top-up, unused balance can be refunded minus what was consumed and the payment channel's fees; bounty and promotional credit are not refundable. There is also a referral bounty of 18% at level 1 and 6% at level 2 of what referred members pay, which can be withdrawn from $11.11 or moved into the wallet.

Claude Code on one side, your own stack on the other

Client rules are firm, and they are the second half of a sensible split. Claude Fable 5 works only from the official Claude Code and Claude Desktop clients, with Claude Code 2.1.285 or newer. Cursor, Cherry Studio, Cline, OpenCode, any SDK and cURL do not work with Claude lanes; those clients pair only with GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax and Hunyuan lanes, which makes GPT-5.6 Sol the natural pick when your day is spent inside an editor or a script.

Gemini runs on its native protocol only, through Gemini CLI, using :generateContent for single replies and :streamGenerateContent?alt=sse for streams, and it cannot go through /v1/chat/completions. Gemini 3.1 Pro is priced at $1.54 input, $8.94 output, $1.54 cache write and $0.1542 cache read per 1 million tokens. Other lanes such as Qwen3.7 Plus ($0.15 input, $0.4486 output, $0.0444 cache read) always draw on the wallet.

Every request can be traced in the usage log. Relay365 gives you API keys with tags, a key bound to a lane, quick-wiring snippets, and a usage log of the latest 200 requests showing the time, model, input, output, cache recall, charge and result of each call. The log does not show which key made the call, so keep your own naming tidy. There is also a ledger of wallet, bounty and pass accounts, in-site messages to the desk once you log in, and six interface languages.

Duty passes for a heavy Claude or GPT stretch

When a project week is dense, the Relay365 duty desk offers duty passes for Claude and GPT only. A pass comes with a dedicated key that never touches the wallet, and other models always use the wallet. The Claude 14-day pass costs $50.97, draws up to $390 of quota at the duty-pass rate x2.5 — at most about $156 of usage at list prices, and less on models whose pass rate sits above list — and allows at most $210 of quota per 7 days.

The GPT 30-day pass costs $81.53, draws up to $2350 of quota at the duty-pass rate x2.5 (at most about $940 of usage at list prices, less on models whose pass rate sits above list), and allows at most $700 of quota per 7 days. For a mixed workload, the pairing is straightforward: put Claude Fable 5 behind Claude Code on a Claude pass, and keep GPT-5.6 Sol on the wallet for the editor, the SDK and cURL work that the Claude clients cannot accept.

A few pass mechanics are worth reading twice. A call already let through on a pass is drawn in full from the pass even if it runs past the quota or the 7-day ceiling, and the overrun counts against the earliest-expiring card. When a pass is nearly used up, each running call holds $3 x the draw rate and an extra parallel call gets HTTP 429 until the running ones finish. That is the desk protecting the card, not a hidden fee.

Where Relay365 is the wrong fit

Honesty at the desk means naming the cases where Relay365 does not suit you. If your policy requires support, terms or billing handled directly by the model vendor, an independent third-party relay is the wrong instrument, and no rate card changes that. If your workflow needs Claude Fable 5 inside Cursor, Cherry Studio, Cline, OpenCode, an SDK or cURL, this setup will not work for you either, because those clients pair only with GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax and Hunyuan lanes.

Relay365 also does not suit a team that wants one OpenAI-compatible endpoint for every vendor family including Gemini, since Gemini runs on its native protocol only through Gemini CLI. It does not suit anyone who wants a monthly subscription or a single monthly invoice, because the wallet is prepaid with no monthly fee. And if you plan to treat bounty or promotional credit as cash, note that it is not refundable, while unused balance from a top-up can be refunded within 14 days minus what was consumed and the payment channel's fees.

For everyone else, the case is straightforward: one table of lanes, one log you can read line by line, four token types priced separately, and a first top-up you can divide between Claude Fable 5 and GPT-5.6 Sol with the arithmetic in front of you. Create a Relay365 account, add your first balance, and let the duty desk show you exactly where every token went.