Gemini 2.5 Flash and GPT-5.6 Sol pricing on Relay365

One workload, two lanes, one pricing table

At the Relay365 duty desk, we like to keep the arithmetic visible. Suppose you have one production workload and want to send part of it to Gemini 2.5 Flash and part to GPT-5.6 Sol. On Relay365 both lanes sit on the same table, so you can read the rates side by side instead of stitching together separate billing pages. Relay365 is an independent third-party relay, and every rate below is per 1 million tokens in approximate USD, while the account itself is kept in CNY.

Four token types are priced separately on Relay365: input, output, cache write and cache read. That separation matters the moment a workload is split across two lanes, because the mix of prompt tokens, generated tokens and cached tokens is what actually moves the bill. The lines that follow are a worked example, not a quote and not a forecast: for example, if you send 500,000 input tokens, 100,000 output tokens and 250,000 cache read tokens per day on each lane, here is what the 30-day arithmetic looks like at Relay365 rates.

Two more details keep the example honest. First, Gemini 2.5 Flash and GPT-5.6 Sol are different lanes with different rates, so the split itself is a decision you make, and the bill simply follows your traffic. Second, because every request can be traced in the Relay365 usage log, you can check these assumptions against your own calls after the fact rather than trusting a spreadsheet forever.

The Gemini 2.5 Flash line items

Gemini 2.5 Flash on Relay365 is listed at $0.2389 per 1 million input tokens, $1.91 per 1 million output tokens, $0.2389 per 1 million cache write tokens and $0.0236 per 1 million cache read tokens. Here is the example volume worked out line by line.

• Input: 500,000 tokens a day is 0.5 of a million, so 0.5 × $0.2389 = $0.11945 a day; over 30 days that is 15 million tokens × $0.2389 = $3.5835.
• Output: 100,000 tokens a day is 0.1 of a million, so 0.1 × $1.91 = $0.191 a day; over 30 days that is 3 million tokens × $1.91 = $5.73.
• Cache read: 250,000 tokens a day is 0.25 of a million, so 0.25 × $0.0236 = $0.0059 a day; over 30 days that is 7.5 million tokens × $0.0236 = $0.177.

Add the daily lines together and the Gemini side of this example runs at $0.11945 + $0.191 + $0.0059 = $0.31635 a day. Multiply by 30 days and you get $0.31635 × 30 = $9.4905 for the month on the Gemini 2.5 Flash lane.

Cache write is listed for Gemini 2.5 Flash at the same $0.2389 rate as input, so if your workload also writes cache, those tokens can be added at that rate on top of the example. No official list price is quoted for this family, so the duty desk makes no comparison for Gemini 2.5 Flash — the Relay365 table simply shows the rate you pay.

The GPT-5.6 Sol line items

GPT-5.6 Sol on Relay365 is listed at $0.3736 per 1 million input tokens, $2.09 per 1 million output tokens and $0.0292 per 1 million cache read tokens. No cache write rate is listed for this model, so the example below carries only the three priced token types.

• Input: 0.5 × $0.3736 = $0.1868 a day; over 30 days, 15 million tokens × $0.3736 = $5.604.
• Output: 0.1 × $2.09 = $0.209 a day; over 30 days, 3 million tokens × $2.09 = $6.27.
• Cache read: 0.25 × $0.0292 = $0.0073 a day; over 30 days, 7.5 million tokens × $0.0292 = $0.219.

The daily sum is $0.1868 + $0.209 + $0.0073 = $0.4031, and the 30-day sum is $0.4031 × 30 = $12.093 for the GPT-5.6 Sol lane in this example.

For comparison, the official list price for GPT-5.6 Sol is input $4, output $20 and cache read $0.4, and the Relay365 output rate is about 10% of official. That is the only comparison drawn here, because the pricing table only supports a comparison where an official list price is actually quoted.

Adding the two lanes, then choosing wallet or duty pass

Put the two lanes side by side for the whole example: $9.4905 on Gemini 2.5 Flash plus $12.093 on GPT-5.6 Sol equals $21.5835, so roughly $21.58 for 30 days of the split workload. If you had sent the same full volume to a single lane instead, the example would come to $18.981 on Gemini 2.5 Flash and $24.186 on GPT-5.6 Sol, so the split in this example lands between the two single-lane totals.

Relay365 bills from a prepaid wallet with no monthly fee, and every call is settled on the token counts the upstream reports. A call that fails upstream with a 502 or 503 is not billed, and a disconnected stream is billed for what was actually used. That is the practical reason the four token types are shown separately on the table: the usage log records input, output and cache recall for each request, so the bill and the log can be read against each other.

For heavier GPT traffic, the duty desk can issue a GPT duty pass, and it comes with a dedicated key that never touches the wallet. One option is $81.53 for 30 days, drawing up to $2350 of quota at the duty-pass rate x2.5, which is at most about $940 of usage at list prices, with at most $700 of quota per 7 days. A larger option is $194.03 for 30 days, drawing up to $5650 of quota at the duty-pass rate x2.5, at most about $2260 of usage at list prices, with at most $1640 of quota per 7 days. At the example volume above, the $12.093 wallet bill for the GPT lane is smaller than the $81.53 pass price, so a pass earns its place only when GPT usage climbs well past this example.

Two pass mechanics are worth knowing before you switch a key over. 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. And when a pass is nearly used up, each running call holds $3 x the draw rate while an extra parallel call gets HTTP 429 until the running ones finish. Gemini always uses the wallet, so a split workload keeps its Gemini traffic on wallet billing while GPT traffic can move to a pass.

On the money side, a new Relay365 account starts with $0.25 in signup credit. The referral bounty pays 18% at level 1 and 6% at level 2 of what referred members pay, withdrawable from $11.11 or moved into the wallet. Within 14 days of a top-up, unused balance can be refunded minus what was consumed and the payment channel's fees, while bounty and promotional credit are not refundable.

Tracing the workload on the Relay365 dashboard

Every lane sits on one table and every request can be traced in the usage log, which shows the latest 200 requests with the time, model, input, output, cache recall, charge and result of each call. The log does not show which key made the call, so if you want per-key clarity, the dashboard lets you create API keys with tags and bind a key to a lane. Quick-wiring snippets are there to get the first call out, and a ledger tracks the wallet, bounty and pass accounts side by side. You can also send in-site messages to the desk after logging in, and the interface is available in six languages.

Client rules are firm, and they differ by lane. Gemini runs on its native protocol only, through Gemini CLI: :generateContent for single replies and :streamGenerateContent?alt=sse for streams, and it cannot go through /v1/chat/completions. The GPT-5.6 Sol lane is the one that pairs with Cursor, Cherry Studio, Cline, OpenCode, SDKs and cURL, which keeps a two-lane setup practical when your team works across different tools.

There are two traffic limits to keep in mind on Relay365: more than 120 calls in 60 seconds on one key, or more than 8 requests in flight for one account, and either can return HTTP 429. For the example workload above, neither limit is anywhere near being the deciding factor, which is another reason to start with a small split and read the usage log before scaling it up.

That is the whole desk procedure: pick two lanes, put the same token volumes on each, and let the per-token rates do the talking. If you would like to run the arithmetic on your own traffic, create a Relay365 account and point your workload at the lanes you want to compare.