Qwen 3.7 Plus Pricing

Qwen 3.7 Plus API pricing

Qwen 3.7 Plus API pricing is $0.32 per million input tokens and $1.28 per million output tokens, verified September 14, 2026.

Input (per 1M tokens)

$0.32

Output (per 1M tokens)

$1.28

Released

June 2, 2026

Capabilities

Qwen 3.7 Plus supports image input, tool use.

Qwen 3.7 Plus API pricing has changed 1 time on record, most recently on July 20, 2026 (decrease).

• July 20, 2026: Decrease — input $0.32/M — output $1.28/M

Frequently asked questions

Short answers to the most common questions about Qwen 3.7 Plus pricing, limits, and capabilities.

Notes

Multimodal sibling to Qwen 3.7 Max, positioned as the mid-tier workhorse for image + video input use cases. Released June 2, 2026. Priced $0.40/$1.60 per MTok input/output — significantly cheaper than Max ($1.25/$3.75 promo). Available via Alibaba Model Studio (Cloud) and OpenRouter. [2026-07-20] Input fell $0.40 → $0.32 and Output fell $1.60 → $1.28 (−20%) per Alibaba Cloud Model Studio international endpoint pricing (USD, ap-southeast-1), corroborated by OpenRouter/pricepertoken. Values are the base input bracket — Qwen tiers by input length; long-context requests bill up to $0.96/MTok input. Mainland (¥) endpoint prices differ. [2026-08-31] Verification attempt produced no clear result — recheck next run. Sources this run disagree on the base bracket: one reports $0.40 input / $1.20 output, another describes tiered input brackets with $0.32 for short inputs escalating to $0.96 for long-context, and neither reproduces the $1.28 output currently stored. Prices LEFT UNCHANGED at $0.32/$1.28 pending a direct read of the Alibaba Model Studio international (ap-southeast-1) USD price list. Complicating factor: Model Studio quotes region-specific rates and the Beijing endpoint runs 60-70% cheaper, so aggregator snippets are unreliable for this row. [2026-09-14] Verification attempt produced no clear result — recheck next run. Second consecutive week of the same three-way split, now with a fourth variant: one source gives $0.40 input / $1.20 output, VentureBeat gives $0.40/$1.60 with $0.08 cached, and a third again describes tiered input brackets ($0.32 short input rising to $0.96 long-context) and adds cache detail — Cache Read $0.064/MTok, Cache Creation 5min $0.40/MTok, Cache Read 5min $0.032/MTok. Still no source anywhere reproduces the stored $1.28 output. Prices LEFT UNCHANGED at $0.32/$1.28. Assessment: this row is not a stale-price problem, it is a modelling-convention problem — Qwen bills by input-length bracket and by region, and the schema has one slot for each field, so no single published number is 'correct' until the convention is fixed. ACTION FOR KHALED: decide which bracket and which endpoint the calculator should represent (suggest: international ap-southeast-1, short-input bracket), record that decision on this row, and the weekly run can then verify against a single target instead of re-flagging the split every week.