Skip to main content
AI Choice Engine

Model comparison

Qwen 3.8 27B vs Qwen 3.8 Max

Two Qwen tiers compared on the figures that decide which one a workload actually needs.

Qwen 3.8 27B: profile checked October 25, 2026Qwen 3.8 Max: profile checked October 25, 2026Individual provider fields may changeEvidence confidence: Medium — see receipts below

Qwen

Qwen 3.8 27B

Budget · Open weights

vs

Qwen

Qwen 3.8 Max

Frontier

AI model capability comparison
SpecificationQwen 3.8 27BQwen 3.8 Max
ProviderQwenQwen
TierBudgetFrontier
Context window262KWinner: 1M
Max outputNot verifiedUnverified131K
Input / 1M tokensWinner: $0.50$2
Output / 1M tokensWinner: $3$6
WeightsOpenClosed
Parameters27B dense VLM (Gated DeltaNet hybrid)2.4T total (MoE; managed active count not verified)
Reasoning levelslow, medium, xhighlow, medium, xhigh
Modalitiestext, image, videotext, image, video
LicenseApache 2.0Not verifiedUnverified
API model idqwen3.8-27bqwen3.8-max
ReleasedNot verifiedUnverifiedAugust 3, 2026
Artificial Analysis Intelligence Index [xhigh] · source date Qwen 3.8 27B: 2026-10-0134Not verifiedUnverified
Artificial Analysis Intelligence Index · source date Qwen 3.8 Max: 2026-10-01Not verifiedUnverified45
Terminal-Bench 2.1 · source date Qwen 3.8 Max: 2026-10-01Not verifiedUnverified86.6
GPQA Diamond · source date Qwen 3.8 Max: 2026-10-01Not verifiedUnverified92.6
Humanity's Last Exam · source date Qwen 3.8 Max: 2026-08-14Not verifiedUnverified56.2

Dates beside benchmark scores refer to the linked source’s publication or capture date unless that source explicitly states when it measured the result. A shared URL alone does not establish comparable runs; unverified setups are marked. Prices are USD per million tokens at standard rates, excluding batch and caching discounts. Values we could not confirm from the provider are shown as “Not verified” rather than estimated.

Benchmark receipts

Where each score comes from, and how far it can be compared across models.

Pricing tiers

Qwen 3.8 27B: Hosted qwen3.8-27b International lists USD/MTok input/output $0.50/$3; Beijing lists $0.424/$1.696. QwenCloud marketplace quotes implicit cache $0.10, explicit creation $0.625/read $0.05. Its hosted card lists 1M context and 131K maximum response; the Apache-2.0 downloadable checkpoint has 262,144 native context. August 19 is the hosted listing, not proof of exact weight-publication day. Recommended generation budgets do not establish a hard native output cap.

Qwen 3.8 Max: Alibaba Model Studio USD/MTok: Singapore International and Virginia US scope input/output $2/$6, implicit cache $0.25, explicit creation $2.50/read $0.17. Beijing and Global scopes $1.65/$4.951, implicit $0.206, explicit creation $2.063/read $0.137. The base Beijing card quotes Batch File $0.825/$2.475 and Batch Chat $1.65/$4.951, but the 0902 snapshot card marks Batch unsupported; do not assume alias or snapshot Batch eligibility from the price cells. Other scopes mark Batch unsupported. These are list prices excluding console-only offers. Context 1,000,000; input 991,808 non-thinking/983,616 thinking; output 131,072 and reasoning 262,144 are separate published limits, with permitted combinations dependent on request settings.

BudgetOpen weightsRecord checked October 25, 2026

Qwen 3.8 27B

Qwen 3.8 27B is Alibaba's compact deployment-friendly dense VLM — 262K native context (1M via YaRN), Apache 2.0 weights, distinct from the hosted Qwen 3.8 Max API.

Best for

  • Self-hosting on a single node
  • Multimodal input at small scale
  • Long-context on modest hardware

Watch out

Native thinking defaults xhigh and supports low/medium/xhigh or non-thinking; preserve_thinking defaults true. Exact checkpoint release day and hard native output cap remain unverified. Managed 1M limits do not transfer to local weights. The native card recommends 131,072 final output for complex work, not a guaranteed generation cap. Quantization, memory and context scaling need target-hardware validation; video is input, text is output.

FrontierRecord checked October 25, 2026

Qwen 3.8 Max

Qwen 3.8 Max is Alibaba's managed multimodal flagship, with a vendor-described 2.4T MoE scale and 1M context.

Best for

  • Hosted agentic work
  • Long-context
  • Multimodal

Watch out

Managed Max accepts text/image/video and thinking on/off; native 2.4T-A95B is text-only and requires thinking. Native active-parameter counts and 262,144 context do not describe the managed serving variant. Low/medium/xhigh, default xhigh, with preserved thinking by default; reasoning_effort and thinking_budget cannot be set together. The September 5 alias update targets 0902, while native checkpoint and historical benchmark identities stay separate. Base-card versus snapshot Batch flags conflict; region/API-key scope must be checked.

When the cheaper one wins

Qwen 3.8 27B is cheaper on output at $3 per million tokens against $6 for Qwen 3.8 Max — about 2.0×. Use the cheaper tier for classification, extraction, summarisation, and any task where the expensive model’s extra score does not change the accepted output. The expensive one only pays if your hardest task actually fails on the cheap tier. This compares standard short-context token rates, not total cost per accepted task. Include input, cache, context tiers, schedules, retries and review in your pilot.

Run the model picker

Evidence confidence: Medium

How strong and complete the evidence behind this comparison is — not a prediction of which model is better.

  • Pricing verified on both sides — Input and output rates are verified for both models.
  • 3/5 core specs verified on both sides — Not published for at least one side: max output, release date.
  • No directly aligned benchmark setup — No benchmark has matching cited sources, source dates, units, effort and recorded run details on both models.
  • Verified within the last 90 days — Older of the two catalog checks was 1 day ago.
  • Both models carry source citations — Each side has at least two catalog sources on record.

Source receipts

Each catalog figure was checked against the provider or an independent second source on the date shown.

Related comparisons

Diving deeper on one model? Qwen 3.8 27B · Qwen 3.8 Max

Common questions

Qwen 3.8 27B vs Qwen 3.8 Max

Answered from the verified figures on this page rather than general guidance.

Is Qwen 3.8 27B or Qwen 3.8 Max cheaper for input?
Qwen 3.8 27B is cheaper at $0.50 per million input tokens, against $2 for Qwen 3.8 Max — roughly 4.0× the price. Estimate both input and output token volume for your workload. These are base rates; Qwen 3.8 27B: Hosted qwen3.8-27b International lists USD/MTok input/output $0.50/$3; Beijing lists $0.424/$1.696. QwenCloud marketplace quotes implicit cache $0.10, explicit creation $0.625/read $0.05. Its hosted card lists 1M context and 131K maximum response; the Apache-2.0 downloadable checkpoint has 262,144 native context. August 19 is the hosted listing, not proof of exact weight-publication day. Recommended generation budgets do not establish a hard native output cap. Qwen 3.8 Max: Alibaba Model Studio USD/MTok: Singapore International and Virginia US scope input/output $2/$6, implicit cache $0.25, explicit creation $2.50/read $0.17. Beijing and Global scopes $1.65/$4.951, implicit $0.206, explicit creation $2.063/read $0.137. The base Beijing card quotes Batch File $0.825/$2.475 and Batch Chat $1.65/$4.951, but the 0902 snapshot card marks Batch unsupported; do not assume alias or snapshot Batch eligibility from the price cells. Other scopes mark Batch unsupported. These are list prices excluding console-only offers. Context 1,000,000; input 991,808 non-thinking/983,616 thinking; output 131,072 and reasoning 262,144 are separate published limits, with permitted combinations dependent on request settings.
Is Qwen 3.8 27B or Qwen 3.8 Max cheaper for output?
Qwen 3.8 27B is cheaper at $3 per million output tokens, against $6 for Qwen 3.8 Max — roughly 2.0× the price. Estimate both input and output token volume for your workload. These are base rates; Qwen 3.8 27B: Hosted qwen3.8-27b International lists USD/MTok input/output $0.50/$3; Beijing lists $0.424/$1.696. QwenCloud marketplace quotes implicit cache $0.10, explicit creation $0.625/read $0.05. Its hosted card lists 1M context and 131K maximum response; the Apache-2.0 downloadable checkpoint has 262,144 native context. August 19 is the hosted listing, not proof of exact weight-publication day. Recommended generation budgets do not establish a hard native output cap. Qwen 3.8 Max: Alibaba Model Studio USD/MTok: Singapore International and Virginia US scope input/output $2/$6, implicit cache $0.25, explicit creation $2.50/read $0.17. Beijing and Global scopes $1.65/$4.951, implicit $0.206, explicit creation $2.063/read $0.137. The base Beijing card quotes Batch File $0.825/$2.475 and Batch Chat $1.65/$4.951, but the 0902 snapshot card marks Batch unsupported; do not assume alias or snapshot Batch eligibility from the price cells. Other scopes mark Batch unsupported. These are list prices excluding console-only offers. Context 1,000,000; input 991,808 non-thinking/983,616 thinking; output 131,072 and reasoning 262,144 are separate published limits, with permitted combinations dependent on request settings.
Which has the larger context window, Qwen 3.8 27B or Qwen 3.8 Max?
Qwen 3.8 Max accepts 1M tokens against 262K for Qwen 3.8 27B. This only matters if you routinely send very long documents or large codebases.
Do Qwen 3.8 27B and Qwen 3.8 Max support the same reasoning levels?
Both expose these named effort settings: "low", "medium", "xhigh". The same labels do not guarantee equal compute budgets or behavior across providers. Compare quality, latency and cost on your own tasks.
Should I use Qwen 3.8 27B or Qwen 3.8 Max?
We classify Qwen 3.8 27B in the budget tier and Qwen 3.8 Max in the frontier tier. Test whether either meets your accuracy, tool-use and latency requirements before choosing on price.
Can I self-host Qwen 3.8 27B or Qwen 3.8 Max?
Qwen 3.8 27B publishes open weights, but self-hosting and third-party provider access depend on the licence, hardware, serving support, and availability. Qwen 3.8 Max is a closed model whose supported access paths are controlled by its provider.

Next step

Choosing between them

The verified figures that separate this pair, computed from the catalog rather than restated boilerplate.

  • Input price: Qwen 3.8 27B costs $0.50 per 1M tokens versus $2 for Qwen 3.8 Max — a 4x difference at the headline tier.
  • Context: Qwen 3.8 Max takes 1M against 262K for Qwen 3.8 27B — only decisive if your prompts approach the smaller window.
  • Deployment: Qwen 3.8 27B publishes weights you can self-host; the other is API-only.
  • Positioning: Qwen 3.8 27B sits in the budget tier, Qwen 3.8 Max in the frontier tier. Compare accuracy, reliability, latency, and cost on representative tasks before choosing.
Run the model pickerEstimate stack costAll model comparisons

Choosing a harness rather than a model — Cursor, Copilot, Claude Code, Windsurf, or Lovable? Compare agentic harnesses · Latest releases.