Y-API vs Together AI
Together AI is another gateway reselling the same vendors’ models, so this is not two labs competing — it is one question. For the 2 models both of us sell, does going through Y-API cost less? Both columns below are cash per 1M input tokens, and Together AI’s side is the cheapest tier we could find on its own price list, which makes every multiple a floor rather than a headline.
| Dimension | Y-API | Together AI |
|---|---|---|
| API protocol | Y-APIOpenAI Chat Completions. Point base_url at https://api.y-api.bestvirtualgoods.com/v1, swap the key, leave the rest of your code alone. | Together AIAn OpenAI-compatible endpoint of its own, on a different base URL and with different model strings — see the row for each model below. |
| Models on one key | Y-API10 model IDs from 6 vendors, one balance, one key. | Together AIIts own published catalog. Of the 10 models we compare against it, 2 are priced there. |
| qwen/qwen3.8-max — input / 1M tokens | Y-API$0.10 in cash | Together AI$2.50 as Qwen3.8-2.4T-A95B — 25× ours (listed under a different name) |
| deepseek/deepseek-v4-flash — input / 1M tokens | Y-API$0.0075 in cash | Together AI$0.14 as DeepSeek V4 Flash 0731 — 18.7× ours (dated snapshot build, not the rolling model) |
| What a top-up converts to | Y-APILimited time: $1 becomes $20 of account credit, then credit is deducted at each model’s own rate. Reverts to 1:10 afterwards. | Together AIYou pay the list price in cash. There is no credit multiplier, and nothing to expire. |
| Cache-hit pricing | Y-APINo cache discount. A repeated prompt is billed exactly like a new one. | Together AINo cached-input rate published for the rows above, so both columns assume every request is a cache miss. If Together AI discounts cache hits off the rate card, the gap narrows. |
| Models it does not price | Y-APIAll 10 are callable on the same key. | Together AIOne model we looked for and could not find (xiaomi/mimo-v2.5); 7 models we have not checked against it yet. |
Against us
Where Together AI costs less
A comparison that only runs one way is an advertisement. Everything below comes from the same price list as the figures above.
We have no cache discount at all
Every request is billed at the full rate whether or not the prompt is identical to the last one. Long fixed system prompts and repeated document context are exactly where that hurts.
Some rows are not the same build on both sides
We price qwen/qwen3.8-max against its Qwen3.8-2.4T-A95B and deepseek/deepseek-v4-flash against its DeepSeek V4 Flash 0731. Matching those by vendor, parameter count and release is our judgement, not their claim. The prices are comparable; the behaviour may not be. Pin the model ID and test it if that matters to you.
Most of the gap is the top-up rate, not the model price
Our catalog rates, read as credit, sit close to what Together AI charges in cash. Almost the whole gap above comes from $1 converting to $20 of credit. At 1:10 the same comparison becomes 12.5× on qwen/qwen3.8-max and 9.33× on deepseek/deepseek-v4-flash.
Together AI sells far more models than we do
Its catalog spans vendors and open-weight models we do not carry at all, and it can route around a provider that is down. Our 10 are a deliberate shortlist; if the model you need is not on it, our price for the others is irrelevant to you.
Fit
Which one to call
Call it through Y-API if
- Your input is mostly fresh, where the gap is 25× on qwen/qwen3.8-max and 18.7× on deepseek/deepseek-v4-flash at the current top-up rate.
- You want one balance and one key across 10 models from 6 vendors instead of an account per vendor.
Go to Together AI if
- You need a model outside our 10, or you want automatic fallback across providers when one is down.
- You want a price that does not depend on a promotion. Ours is $1 → $20 of credit today and 1:10 afterwards; theirs is cash at list price either way.
- You want the billing relationship, the SLA and the support to be with Together AI rather than with a reseller.
FAQ
Questions people ask before switching
- Is Y-API cheaper than Together AI?
- On fresh input, yes: qwen/qwen3.8-max runs $0.10 against their $2.50 and deepseek/deepseek-v4-flash runs $0.0075 against their $0.14. Two things bound that. Their side is always the cheapest tier they publish, so each multiple is a floor rather than a peak; and the gap depends on the limited-time 1:20 top-up rate, which roughly halves it when it reverts to 1:10.
- Why is it cheaper — are you buying below their price?
- No, and the honest answer matters here. Our catalog rates, read as credit, sit close to Together AI’s cash list prices. Almost the whole gap is the top-up conversion: $1 buys $20 of credit while the promotion runs. Treat it as a promotional rate with an announced revert to 1:10, not as a structural discount.
- Which models does Together AI price against ours?
- It prices 2 of our 10, with the name each one carries there: qwen/qwen3.8-max (Qwen3.8-2.4T-A95B there) and deepseek/deepseek-v4-flash (DeepSeek V4 Flash 0731 there). Those are the only rows in the table above — a row with a published price on one side only is not a comparison.
- What about the models it does not price?
- One model we looked for and could not find (xiaomi/mimo-v2.5); 7 models we have not checked against it yet. We keep “we did not look” and “they do not sell it” apart on purpose: they are different claims, and only the first one is ours to make.
- Is the model on their side literally the same build?
- Not for qwen/qwen3.8-max (Qwen3.8-2.4T-A95B there) and deepseek/deepseek-v4-flash (DeepSeek V4 Flash 0731 there) — a dated snapshot or a differently named build, matched by vendor, parameter count and release date. Compare the prices; do not assume identical behaviour.
- How much of my code has to change?
- base_url, the key, and the model string. Both sides speak OpenAI Chat Completions, so nothing else in the request body changes — ours are prefixed, for example qwen/qwen3.8-max.
Provenance
Sources
Every Together AI figure on this page was read from the page below on 2026-08-26, at the cheapest tier it publishes. Our own figures are on our pricing and terms pages.
Point one call at it
Change base_url, the key, and the model string on a single call you already make. The request log will tell you what it costs before you move anything else.