
Together cut Kimi K3 prices and raised Qwen3.8 Flash rates
By Rob Litterst, Co-founder, PricingSaaS
Published
Together lowered Kimi K3 token prices and raised Qwen3.8 Flash's token prices.
Together’s pricing page · Change detected
What changed
Together cut Kimi K3's serverless token price 10% to $2.70 per 1M input tokens and $13.50 per 1M output tokens, down from $3.00 and $15.00, according to Together's pricing page on Oct 6, 2026. In the same update, Together raised Qwen3.8 Flash's price to $0.15 input / $0.47 output per 1M tokens, up from $0.09/$0.28, per Together's pricing page on Oct 6, 2026.
Qwen3.8 Flash's new rate matches where it stood before Together's Sep 23, 2026 cut to $0.09/$0.28, according to Together's pricing page — a reversal back to that earlier rate rather than a new high.
Why it matters
For teams buying inference by the token, this is a reminder that a price decrease and a price increase can land in the same vendor update on different models — the net effect on a mixed workload depends entirely on which model carries the volume. A team that shifted traffic from Qwen3.8 Flash to Kimi K3 after the Sep 23 cut is now paying more on the Flash side again, while anyone still on Kimi K3 gets a straightforward 10% saving on both input and output tokens as of Oct 6, 2026, per Together's pricing page.
This also argues for pricing your own usage-based product at the model or SKU level rather than one blanket rate: Together is treating each serverless model as its own lever, moving some down and some up within the same update. Watch whether Qwen3.8 Flash's rate holds at $0.15/$0.47 or moves again — a model that has already changed price twice between Sep 23, 2026 and Oct 6, 2026 is one to monitor closely before locking a customer-facing rate card to it.
In context
Pulse has tracked 40 real pricing changes on Together since Dec 29, 2025, 22 of them in the last 90 days. This lands a day after Together added the free-output Tev1 4B Experimental model on Oct 5, 2026. The Kimi K3 cut sits inside credits-and-usage-billing repricing, a common pattern among AI infrastructure vendors, with names like Lumaai, Katteb and Invideo among the most frequent. For a product leader pricing a token-based API, pairing a real cut on one model with a reversal on another in the same update is a reminder that Together retunes its serverless catalog per model, not as one page-wide number.
Common questions
Does Together say why it reversed Qwen3.8 Flash's price?
No — Together's pricing page lists the new $0.15/$0.47 per 1M token rate on Oct 6, 2026 without an explanation, and Pulse found no company statement to quote, so this briefing runs without a rationale section.
How can I estimate what the Kimi K3 cut saves my own workload?
Apply the same 10% cut Together made to Kimi K3 on Oct 6, 2026, according to Together's pricing page, to your own monthly input and output token counts — since output tokens are priced far above input on Kimi K3, the dollar savings will be larger wherever your workload leans toward output.
Should we expect more price changes to these models soon?
Together has made 40 real pricing changes since Pulse started tracking it on Dec 29, 2025, including 22 in the last 90 days, so further catalog adjustments are plausible — but Pulse has no specific signal pointing to another move on Kimi K3 or Qwen3.8 Flash.
Sources
- Pulse: Together pricing changes, Oct 6, 2026pricingsaas.com/companies/together/diffs/2026D279
- Together pricing pagetogether.ai/pricing