All times Eastern (ET)Provider pages checked daily at
ProSign inRSS

Qwen3.8 Max

API ID qwen3.8-max

qwen3.8-max is a Qwen model. It has a 1,000,000-token context window and up to 131,072 output tokens. It takes image input and supports function calling and structured outputs. Standard price: $2.00 input and $6.00 output per 1M tokens.

Status
Active
Released
—
Shutdown
none announced
Replacement
—

Qwen announces retirement dates without naming the models on a page AIWire can read. AIWire posts each new notice on the wire and will flag qwen3.8-max if it disappears from Qwen's model list.

Benchmarks

Scores as each source published them
or

Comparing with Qwen3.7 Max, the previous model in this line. A score is compared only with the same benchmark from the same source, at the same setting where both have it. Full comparison: prices, limits, features

Arena

ArenaQwen3.8 MaxQwen3.7 MaxChangeRankDateSource
Text
By category
  • Math 1500 rank 18 vs not rated
  • Creative writing 1471 rank 16 vs not rated
  • Instruction following 1474 rank 28 vs not rated
  • Hard prompts 1503 rank 28 vs not rated
  • Long prompts 1492 rank 27 vs not rated
  • Multi-turn 1491 rank 20 vs not rated
  • Expert 1513 rank 23 vs not rated
1482 ±5not rated24 of 415
26,991 votes		Oct 13, 2026source
Text: Coding1523 ±8not rated30 of 410
7,497 votes		Oct 13, 2026source
WebDev1672 ±12not rated11 of 142
3,448 votes		Oct 13, 2026source
Vision1301 ±7not rated2 of 165
10,017 votes		Oct 9, 2026source
Search1229 ±13not rated5 of 42
1,870 votes		Oct 13, 2026source
Agent0.019 ±0.006-0.058 ±0.012+0.07726 of 54
67,520 sessions		Oct 14, 2026source

Rated as qwen3.8-max. Text, WebDev, Vision and Search ratings come from people voting between two anonymous answers; Text, Vision and Search are style-controlled, as on arena.ai. Agent is Arena's separate score for real agent sessions, on its own scale. Data: Arena (arena.ai), CC BY 4.0.

Epoch AI

BenchmarkQwen3.8 MaxQwen3.7 MaxChangeSettingRunSource
FrontierMath Tier 446.3%34.1%+12.2 ptsxhigh effort (vs no setting)Aug 4, 2026source
FrontierMath Tiers 1-374.7%64.6%+10.1 ptsxhigh effort (vs no setting)Aug 4, 2026source
GPQA Diamond92.7%90.9%+1.8 ptsxhigh effort (vs no setting)Aug 4, 2026source
OTIS Mock AIME 2024-202599.4%95.6%+3.8 ptsxhigh effort (vs no setting)Aug 4, 2026source
SimpleQA Verified45.8%55.8%−10.0 ptsxhigh effort (vs no setting)Aug 27, 2026source

Run independently by Epoch AI. Data: Epoch AI, Capabilities & benchmarking (epoch.ai/benchmarks), CC BY.

Features

Compare models
FeatureValueSinceSource
Structured outputsThe model can be held to a JSON schema you supply, so every reply parses without retries or cleanup.Supportedin place when tracking began (October 6, 2026)source
Function callingAlso called tool use. You describe functions; the model replies with a function name and arguments for your code to run, then uses the result.Supportedin place when tracking began (October 6, 2026)source
Prompt cachingA prompt prefix you send again and again (system prompt, documents, tools) is cached, so later requests cost less and start faster.Supportedin place when tracking began (October 6, 2026)source
Batch APISend many requests as one job that runs in the background, usually finished within 24 hours, at a lower price than live requests.Not supported (supported in the Beijing region only)first seen October 8, 2026source
Image inputImages such as screenshots, photos and charts can go in the prompt for the model to read.Supportedin place when tracking began (October 6, 2026)source
Fine-tuningYou can train your own version of the model on your examples through the company's API, then call it like any other model.Not supportedin place when tracking began (October 6, 2026)source
Built-in web searchThe company's API can search the web for the model during a request and cite its sources; you don't need a search service of your own.Supportedin place when tracking began (October 6, 2026)source
Context windowThe most the model can work with in one request, in tokens: your prompt, conversation history and files. A token is about three-quarters of an English word.1,000,000 tokensin place when tracking began (October 6, 2026)source
Max output tokensThe longest reply the model can write in one response, in tokens. A token is about three-quarters of an English word.131,072 tokensin place when tracking began (October 6, 2026)source

Not published by Qwen: PDF input, Reasoning / thinking, temperature parameter, top_p parameter, logprobs.

Price history

USD per 1M tokens, standard tier
$0.00$2.00$4.00$6.00$8.00Oct 2026In $2.00Out $6.00Oct 5, 2026 to Oct 15, 2026: Input $2.00, Output $6.00
$0.00$2.00$4.00$6.00$8.00Oct 2026In $2.00Out $6.00Oct 5, 2026 to Oct 15, 2026: Input $2.00, Output $6.00

We started recording this model's prices on Oct 5, 2026. The history grows with each daily check.

Show as a table
FromTypePrice
Oct 5, 2026Input$2.00
Oct 5, 2026Output$6.00

Prices

USD per 1M tokens, every type and band, by tier
TypePrice $/1MIn effect
Standard
input$2.00
cached input$0.25
cache write$2.50
output$6.00

Changelog

Changes we detected on Qwen's pages
  1. Feature changed

    qwen3.8-max: Batch API not supported → not supported (supported in the Beijing region only)

    Info

Email alerts

Get an email if qwen3.8-max changes

We'll email you the same morning we spot a deprecation notice, a new or moved shutdown date, a price change, or the model leaving Qwen's model list.

No password: one email to confirm, and you can unsubscribe any time. Privacy

Last checked · Sources: Qwen models, Qwen deprecations, Qwen free terms