Provider pricing hub

Qwen API pricing: Max, Plus and Flash.

Qwen API prices on RouterPlex run from $0.15 per 1M input tokens (Qwen3.8 Flash) to $2.50 (Qwen3.7 Max), with 0% markup, and every featured Qwen route lists 1M context. Qwen3.8 Max costs less than Qwen3.7 Max. No Alibaba Cloud account needed. Top-ups have a $0 fee, by card from $5 or crypto; balances stop at $0 and each key has a hard USD cap.

Prices checked 17 November 2026 from the live catalog. Alibaba Cloud's pricing page last checked 10 Nov 2026: www.alibabacloud.com/help/en/model-studio/model-pricing

Qwen on this key17 November 2026

8

Routes

$0.15

Cheapest input / 1M

Markup 0%

Qwen family

Published catalog rates.

Agent-turn cost uses 100K input and 10K output. Max, Plus and Flash are Alibaba's tiers, not quality scores.

Model routeContextInput / 1MCached input / 1MOutput / 1MAgent turn
Qwen3.8 Flash

qwen3.8-flash

1M$0.15$0.016$0.47$0.020
Qwen3.6 Flash

qwen3.6-flash

1M$0.25not offered$1.50$0.040
Qwen3.7 Plus

qwen3.7-plus

1M$0.32$0.12$1.28$0.045
Qwen3.6 Plus

qwen3.6-plus

1M$0.50not offered$3.00$0.080
Qwen3.8 Max

qwen3.8-max

1M$2.00$0.70$6.00$0.26
RouterPlex bills $2 input / $6 output per 1M tokens, Alibaba's list price. Cached prompt tokens are billed at $0.70 per 1M, above Alibaba's own cache-hit tier: our cache price is set from what our upstream charges us for a cached token, so it does not follow the lab's rate.
Qwen3.8 Max 0902

qwen3.8-max-0902

1M$2.00$0.70$6.00$0.26
Dated 2026-09-02 snapshot of Qwen3.8 Max, billed exactly like qwen3.8-max.
Qwen4 Max Preview

qwen4-max-preview

1M$2.00not offered$6.00$0.26
Preview model. Alibaba offers Qwen4-Max Preview only on its Token Plan subscription and publishes no per-token API price, so RouterPlex bills Qwen3.8 Max's list rate of $2 input / $6 output per 1M tokens. Alibaba retires the preview on November 25, 2026. Reasoning is on by default, so allow a generous max_tokens or short replies can come back empty.
Qwen3.7 Max

qwen3.7-max

1M$2.50$0.35$7.50$0.33

Qwen pricing calculator

Same workload. Every Qwen route.

Defaults are a 10K-in / 1K-out request, 3,000 times a month. Switch to the 100K / 10K agent turn for a coding loop, and set a cached share if your prompts repeat a long prefix.

Workload calculator

Catalog rates · USD · no markup

Share of each prompt that repeats a recent prefix (system prompt, files) and bills at the cached-input rate. Routes with no cached rate bill it as normal input.

Models in the estimate

4 selected
10K in + 1K out × 3,000 req$100 of tokens
RoutePer requestMonthly$100 buys
Qwen3.8 Flash

Alibaba

$0.0020$5.9150,761 req
Qwen3.6 Flash

Alibaba

$0.0040$12.0025,000 req
Qwen3.7 Plus

Alibaba

$0.0045$13.4422,321 req
Qwen3.8 Max

Alibaba

$0.026$78.003,846 req

Estimates apply live catalog rates, including cached input, and switch to a route's long-context rate when the prompt reaches its threshold. Actual spend follows the tokens your client sends and the model returns. Image-generation routes are omitted because they are not a chat turn.

Alibaba Cloud vs RouterPlex

What Alibaba's own pages say, and what changes here.

Checked 2026-11-10 on www.alibabacloud.com/help/en/model-studio/model-pricing

  • 01Alibaba prices Qwen through Model Studio. Its list prices Qwen3.7 Plus and Qwen3.6 Plus higher once one request passes 256K input tokens; the RouterPlex catalog rows bill one rate at every length.
  • 02New Model Studio users get a free quota, typically 1M tokens per eligible model for 90 days, in the Singapore region only (International deployment scope). It is Alibaba's offer, not a RouterPlex credit.
  • 03On RouterPlex you skip the Alibaba Cloud account: one key, one prepaid balance, the same ids as the table above.

Call it

Use Qwen from Claude Code or the OpenAI SDK.

Claude Code and the Anthropic SDK use https://api.routerplex.com. The OpenAI SDK, Cursor and Cline use https://api.routerplex.com/v1. Same key, same catalog id, same per-key budget. Swap qwen3.8-flash for any id in the table.

Claude Code (shell)
export ANTHROPIC_BASE_URL=https://api.routerplex.com
export ANTHROPIC_AUTH_TOKEN="$ROUTERPLEX_API_KEY"
export ANTHROPIC_MODEL=qwen3.8-flash
claude
OpenAI SDK (Python)
import os
from openai import OpenAI
 
client = OpenAI(
base_url="https://api.routerplex.com/v1",
api_key=os.environ["ROUTERPLEX_API_KEY"],
)
reply = client.chat.completions.create(
model="qwen3.8-flash",
messages=[{"role": "user", "content": "Hello!"}],
)
print(reply.choices[0].message.content)

How to pick

Max, Plus or Flash: cost per turn.

These notes only use catalog fields: price, context, and mode. They are not quality rankings.

Cheapest per agent turn

Qwen3.8 Flash costs about $0.020 for a 100K-in / 10K-out turn at $0.15 / $0.47 per 1M. Use it for short, retryable work.

The newer Max costs less

Qwen3.8 Max is $2.00 / $6.00 per 1M; Qwen3.7 Max is $2.50 / $7.50. Unless your evals prefer 3.7, the newer Max is the cheaper Max.

Plus sits in between

Qwen3.7 Plus is about 2.3x the agent-turn cost of Qwen3.8 Flash. Try it where Flash answers fall short before paying Max rates.

Setup notes for Qwen3.8 Max, Qwen3.8 Flash and the OpenAI-compatible API. Compare the rest of the catalog on all model pricing, the cheapest-first ledger, or RouterPlex vs OpenRouter.

Before you bill production

Qwen pricing, in questions.

01How much does the Qwen API cost?+

As of 17 November 2026, Qwen routes on RouterPlex cost $0.15 to $2.50 per 1M input tokens and $0.47 to $7.50 per 1M output tokens with 0% markup: Qwen3.8 Flash is $0.15 / $0.47, Qwen3.6 Flash is $0.25 / $1.50, Qwen3.7 Plus is $0.32 / $1.28, Qwen3.8 Max is $2.00 / $6.00, Qwen3.8 Max 0902 is $2.00 / $6.00, Qwen4 Max Preview is $2.00 / $6.00 and Qwen3.7 Max is $2.50 / $7.50. Top-ups have a $0 fee and the balance stops at $0.

02Which Qwen model is cheapest?+

Qwen3.8 Flash at $0.15 / $0.47 per 1M tokens, about $0.020 for a 100K-in / 10K-out agent turn. Cheapest means catalog price per turn, not quality; the cross-provider ranking is on /models/cheapest.

03Is Qwen3.8 Max cheaper than Qwen3.7 Max?+

Yes. Qwen3.8 Max is $2.00 / $6.00 per 1M against $2.50 / $7.50 for Qwen3.7 Max, so the newer Max costs less on both input and output: about $0.26 per agent turn against $0.33.

04How do I get a Qwen API key?+

On RouterPlex you do not need a Alibaba Cloud account: sign up, verify your email, and create a key in the dashboard (you can set a hard USD budget on it). Call https://api.routerplex.com/v1 with the OpenAI SDK, or point Claude Code at https://api.routerplex.com, with model "qwen3.8-flash". The same key reaches every other model in the catalog.

05Is there a free Qwen API tier on RouterPlex?+

No. RouterPlex has no free tier. Eligible verified accounts get $1 of test credit once, no card, and it never expires. After that you top up from $5 by card, or with crypto, with a $0 top-up fee; the balance stops at $0, so a Qwen loop cannot run up a bill. Alibaba's own Model Studio gives new users a free quota, typically 1M tokens per eligible model for 90 days, only in the Singapore region with the International deployment scope; it applies to Alibaba accounts, not to a RouterPlex key. (Source: www.alibabacloud.com/help/en/model-studio/new-free-quota, checked 2026-11-10.)

06Which Qwen model id should I use?+

Send the catalog id exactly as written: "qwen3.8-flash" (Qwen3.8 Flash), "qwen3.6-flash" (Qwen3.6 Flash), "qwen3.7-plus" (Qwen3.7 Plus), "qwen3.8-max" (Qwen3.8 Max), "qwen3.8-max-0902" (Qwen3.8 Max 0902), "qwen4-max-preview" (Qwen4 Max Preview) and "qwen3.7-max" (Qwen3.7 Max). The cheapest per agent turn is "qwen3.8-flash". The page URL slug (for example /models/qwen3-8-flash) is not the API id; the id keeps its dots and case. Anything else returns model_not_found.

07How does Qwen cache pricing work?+

Cached input is the part of a prompt the provider has already processed in a recent request with the same prefix, such as a long system prompt or a repeated file. It bills at the cached-input rate instead of the full input rate: $0.016 per 1M on Qwen3.8 Flash (input $0.15), $0.12 per 1M on Qwen3.7 Plus (input $0.32), $0.70 per 1M on Qwen3.8 Max (input $2.00), $0.70 per 1M on Qwen3.8 Max 0902 (input $2.00) and $0.35 per 1M on Qwen3.7 Max (input $2.50). Qwen3.6 Flash and Qwen4 Max Preview list no cached rate, so repeats bill as normal input. Set a cached share in the calculator above to see the effect.

08Do long Qwen prompts cost more?+

On Alibaba's own price list, Qwen3.7 Plus and Qwen3.6 Plus cost more once a single request passes 256K input tokens. The RouterPlex catalog rows for every Qwen route list one rate at every prompt length, so the calculator above uses that single rate. (Source: www.alibabacloud.com/help/en/model-studio/model-pricing, checked 2026-11-10.)

09Can I use Qwen in Claude Code?+

Yes. Set ANTHROPIC_BASE_URL to https://api.routerplex.com, use your RouterPlex key as ANTHROPIC_AUTH_TOKEN, and set ANTHROPIC_MODEL to a catalog id such as "qwen3.8-flash". The OpenAI SDK uses https://api.routerplex.com/v1 with the same key.

Route Qwen from one balance

Try Qwen3.8 Flash with $1, no card.

Try it with $1 of test credit, no card. Guided setup shows the exact Claude Code config for Qwen3.8 Flash.

Eligible verified accounts get the $1 once. It never expires.

Qwen API Pricing (Nov 2026): Max, Plus, Flash · RouterPlex