Qwen4 Max API pricing

✓ Matches the official page (checked 2026-11-19 01:00 UTC) · Official pricing ↗Synced 2026-11-19 01:00 UTC · Method

Qwen4 Max prices per 1M tokens
FieldPer 1M tokens
Input$1.50
Output$4.50
Cached input$0.075
Cache write$1.875
Cache write (1-hour TTL)$3.00

Report a wrong price

Alibaba (Qwen) · Max input 991,808 · Max output 131,072
vision · reasoning · tool calling · structured output · thinking mode · non-thinking mode
Per 1K tokens: $0.0015 input, $0.0045 output.

Calculate my cost →

Follow this price:
Atom feed ▾
AtomJSON

Get notified in any feed reader.

Billing rules for caching, Batch and long prompts: Alibaba (Qwen) API cost guide.

Price in context

Qwen4 Max ranks 7 of 10 current Alibaba (Qwen) models by output price, cheapest first (Qwen Flash is the cheapest at $0.40, Qwen3.7 Max the most expensive at $7.50). Output tokens cost 3× the input price; the median across all 119 current models is 5×. Against the whole table (median $1.25 input / $4.40 output per 1M), Qwen4 Max is pricier on both input and output.

Cached input costs 5% of the normal input price ($0.075 vs $1.50 per 1M). In the support-chatbot scenario with 60% cache hits the month comes to $389 instead of $540 without caching (28% less). Writing a prefix to the cache costs 1.25× the input price ($1.875), so caching pays off once a prefix is reused 2 or more times. No batch discount is published for this model.

The context window is 991,808 tokens, roughly 1,984 pages of English text; 67% of current models accept more. The closest model from another provider by price is GLM-5.3 ($1.40 / $4.40); the cheapest reasoning model for the chatbot scenario is MiMo V2.6 Flash at $27.18 per month against $389 for Qwen4 Max.

What it costs per month

Estimated monthly cost by workload (USD)
Scenario (per month)Qwen4 Max
Customer support chatbot
5,000 req/day · 1,200 in · 400 out · cache 60%
$389
RAG over documents
2,000 req/day · 8,000 in · 600 out · cache 30%
$686
Coding agent
1,500 req/day · 25,000 in · 2,000 out · cache 80%
$852
Deep reasoning
800 req/day · 3,000 in · 6,000 out · cache 20%
$735
Batch processing
50,000 req/day · 1,500 in · 150 out · batch
$4,388

Adjust the numbers in the calculator.

Cheaper alternatives

Lowest monthly cost for the chatbot scenario among reasoning models:

Comparisons

Other Alibaba (Qwen) models

Qwen3.8 Max, Qwen3.7 Max, Qwen Max, Qwen3.8 Flash, Qwen3.7 Plus, Qwen3.5 Plus retiring 2027-01-19, Qwen Flash, Qwen Plus, Qwen Coder · All Alibaba (Qwen) API prices →

Models at a similar price

GLM-5.3 ($1.40/$4.40), GLM-5.2 ($1.40/$4.40), GLM-5.1 ($1.40/$4.40), GLM-4.5 AirX ($1.10/$4.50), Muse Spark 1.3 ($1.25/$4.25), Muse Spark 1.2 ($1.25/$4.25)

Price history

No price changes recorded since tracking began on 2026-10-08. See all price changes →

Other changes (4)

Listings, fields a source added or dropped, and limit changes. These are not price moves.

Recorded price changes, USD per 1M tokens
DateChangeOld /1MNew /1MWhat happened
2026-11-17Cache write—$1.875newly listed
2026-11-17Cache write (1-hour TTL)—$3.00newly listed
2026-11-17Cached input—$0.075newly listed
2026-11-17Added———

Source and checks

Source and checks
Data sourceLiteLLM price list (MIT)
Official pricingwww.alibabacloud.com ↗
Last synced2026-11-19 01:00 UTC
Last price changeNo price changes recorded since tracking began on 2026-10-08.
Cross-checkSingle source: models.dev has no comparable data
Official checkVerified against the official page on 2026-11-19 (automatic check). Thinking and non-thinking mode share one output price
Price regionInternational (Singapore)
Field-by-field check against the official page
FieldThis siteOfficialStatus
Input$1.50$1.50✓ official
Output$4.50$4.50✓ official
Cached input$0.075—– unchecked
Cache write$1.875—– unchecked
Cache write (1-hour TTL)$3.00—– unchecked
Max input991,808—– unchecked
Max output131,072—– unchecked

✓ official = the official page's value is in use (other sources' values noted) · ✓ verified = other source equals the official page · ≠ mismatch = differs and is held for confirmation; both shown · … pending = the automatic check could not read the page · – unchecked = not listed on the official page or not checked yet