Together hikes Qwen3.7-Max pricing again, doubling rates in five days
By Rob Litterst, Co-founder, PricingSaaS
Published
Together’s pricing page · Change detected
What changed
Together raised Qwen3.7-Max serverless inference pricing again: input tokens go from $2.00 to $2.50 per 1M (+25%), output from $6.00 to $7.50 per 1M (+25%). Cached input holds flat at $0.25 per 1M. This is the second increase on this model in five days — input was $1.25 and output $3.75 as recently as September 11 — so both rates have now exactly doubled over that span.
Why it matters
Customers running workloads on Qwen3.7-Max have now seen input and output costs double in under a week, with no signal that pricing has stabilized. Watch whether Together lets the new $2.50/$7.50 rate hold or keeps stepping it up as utilization grows, and whether other providers hosting the same open model start closing the gap rather than competing on price.
In context
This is Together's second Qwen3.7-Max price hike in five days, after the 60% jump on September 11, when input rose from $1.25 to $2.00 per 1M tokens, and its third pricing move since Sunday's hike and DeepSeek V4.1 Flash launch. Together has changed pricing 19 times in the last 90 days, but this run is unusually concentrated on one model rather than the usual mix of additions and removals. Price increases remain a minority pattern industry-wide, appearing in just 3.5% of the 5,512 changes tracked over six months, and Together's three hikes in that window already place it among the more aggressive repricers alongside X-Ways and Microsoft. Doubling Qwen3.7-Max's rates in under a week looks less like a correction and more like Together testing what the market will bear for its most in-demand model.