Together raises Kimi K3 cached-input price to $0.30 per 1M tokens
By Rob Litterst, Co-founder, PricingSaaS
Published Last updated
Together’s pricing page · Change detected
What changed
In the serverless inference pricing table, Together raised the Kimi K3 cached input rate from $0.27 to $0.30 per 1M tokens. Standard input ($3.00) and output ($15.00) per 1M tokens are unchanged, and the bump applies only to the discounted cached-token rate.
Why it matters
Teams running Kimi K3 with heavy prompt caching now pay $0.30 per 1M cached input tokens, 11% more than yesterday's listed $0.27, while standard input and output stay at $3.00 and $15.00. If $0.27 was the promo rate, Kimi K3 is now back at full list price on all three lines. Watch whether Together runs another Kimi K3 promo or moves the standard rates next.
In context
Together first showed a single 10%-off Kimi K3 promo ($2.70 input / $13.50 output, from $3.00 / $15.00, valid until Oct 11) in the Oct 6 capture; the week-42 roll-up covers that same promo rather than a second one. Standard rates were back at $3.00 / $15.00 by Oct 13, and the cached-input rate now moves from $0.27 to $0.30, an 11% increase that undoes a 10% discount. Together's previous change was on Oct 8, cutting the GPU Cluster minimum reservation from 7 days to 3, and it averages a change roughly every 8 days (34 in the last 12 months). Elsewhere, ElevenLabs pulled its Eleven v4 credits promo the same day, and Mintlify swapped free credits for a trial a day earlier.