Explore ProductsAI WorkspaceEnterpriseApps & SystemsManaged agentsWhite labelAPI
PlatformsDesktop AppChrome ExtensionWeb App
Explore ResourcesResourcesUse casesCase studies
Mission
SPIRITT logoSPIRITTQwenQwen3.8-Max

Qwen3.8-Max turns 2.4T scale into long agent work

Qwen built its flagship for sustained coding, research, tool use, and professional work. Artificial Analysis scores it at 58 on the Intelligence Index. SPIRITT serves it through U.S.-operated infrastructure, independently of Qwen's own cloud, with 262K of text context, function calling, and $2/M input plus $6/M output reference pricing.

Try Qwen3.8-Max Now
See how it works

Qwen3.8-Max across X

The official release, blind preference data, hands-on visual work, independent cost evidence, and long-horizon agent reactions.

A long-horizon flagship with one route caveat

The native model and the exact SPIRITT service are related, but they do not expose the same capability contract.

What Qwen launched and what SPIRITT serves

Qwen3.8-Max is Qwen's 2.4-trillion-parameter flagship, with roughly 95B active parameters and an official focus on coding, research, visual work, and multi-day agents. The model developer's own service exposes the full native 1M multimodal contract.

SPIRITT serves Qwen3.8-Max through U.S.-operated infrastructure, independently of Qwen's own cloud. The current route offers 262K text-only context, function calling, and reference pricing of $2/M input, $0.25/M cached input, and $6/M output. This page keeps that hosted contract explicit instead of borrowing capabilities from a different route.

Artificial Analysis scores Qwen3.8-Max at 58, with about 21 output tokens per second and roughly $0.91 per Intelligence task in the August 29 snapshot. That is a strong challenger profile, not a universal first-place result.

2.4T total parameters95B active262K SPIRITT contextText + toolsLong-horizon agents$2 / $6 per M

AA Intelligence Index v4.1.1

Claude Opus 5 max
63
Claude Fable 5 max
62
Grok 4.6 high
61
Qwen3.8-Max
58
GLM 5.3 Flash max
57
Qwen3.8 27B xhigh
52

Artificial Analysis model snapshots, August 29, 2026. Qwen3.8-Max scores 58: competitive with the frontier, but behind the current leaders.

Where Qwen3.8-Max is strong, and where the cost shows

Independent rows establish its market position. Qwen's own launch table then shows specific strengths without turning them into an overall winner claim.

Independent knowledge work

GDPval-AA v2 Elo

Claude Opus 5 max
1849
Claude Opus 5 xhigh
1817
Grok 4.6 high
1753
Claude Fable 5 max
1741
Qwen3.8-Max
1737

Artificial Analysis same-harness result. Qwen is in the leading cluster, but behind the displayed Opus, Grok, and Fable variants.

Independent tool use
Winner

Tau3-Banking

Qwen3.8-Max
51.3%
Grok 4.6 high
50.7%

Artificial Analysis multi-turn banking tool-use evaluation. Winner refers only to the displayed pair.

Vendor scientific agents
Winner

PaperBench

Qwen3.8-Max
93%
GPT-5.6 Sol
90.5%
Claude Fable 5
88.8%
Claude Opus 4.8
80.3%

Qwen launch evaluation. Qwen3.8-Max leads the displayed vendor cohort; this is not an independent rerun.

Vendor instruction following
Winner

IFBench

Qwen3.8-Max
82.8%
Qwen3.7-Max
79.1%
GPT-5.6 Sol
72.7%
Claude Fable 5
63.5%

Qwen launch table. Strong instruction following is one of the clearest first-party wins.

Vendor computer use
Winner

OSWorld-Verified

Qwen3.8-Max
86.1%
Claude Fable 5
85%
Claude Opus 4.8
83.4%
Gemini 3.1 Pro
76.2%

Qwen launch table. Agent setup and environment configuration can materially move OSWorld results.

Independent throughput

Output tokens / second

Grok 4.6 high
65.5
Qwen3.8 27B xhigh
50.4
GLM 5.3 Flash max
49.4
Qwen3.8-Max
21

Artificial Analysis model snapshots, August 29, 2026. Qwen3.8-Max is the slowest displayed stream despite its strong quality score.

Hosted economics

Output price ($ / M tokens)

GLM 5.3 Flash
$0.50
Qwen3.8 27B
$3.20
Qwen3.8-Max
$6.00
Kimi K3
$15.00

Reference list prices for the exact hosted routes discussed on these pages. Lower is better; Max buys more capability at a higher operating price.

Independent sources: Artificial Analysis live Qwen3.8-Max model page and same-harness benchmark views, observed August 29, 2026. Vendor sources: Qwen's August 3 launch report and model catalog. SPIRITT route facts were checked against current hosting documentation and the live model route. Vendor benchmark rows use Qwen's harnesses and settings; run your own repeated evaluation before changing a production route.

How It Works

From a bounded objective to Qwen3.8-Max working across files, tools, and verification inside a SPIRITT workspace

01

Give it a long, verifiable objective

Start with a repository, research question, or multi-part professional deliverable. State the constraints, source material, and acceptance checks so a long run has a real finish line.

Open a SPIRITT workspace
02

Pick Qwen3.8-Max

Choose Qwen3.8-Max in the SPIRITT model picker. The SPIRITT route is text-only with 262K context, so bring files and tool results into the workspace rather than assuming the model developer's separate vision contract.

Select Qwen3.8-Max in the model picker
03

Keep the feedback loop in one workspace

Let the model inspect files, call tools, run commands, and verify intermediate results. Keep human approval on high-impact changes and evidence checks on factual claims.

Build and automate with Qwen3.8-Max

Give Qwen3.8-Max a finish line, not a one-line prompt

Run Qwen3.8-Max inside a workspace with files, browser, terminal, tools, memory, and explicit verification.

Try Qwen3.8-Max Now
Official Qwen launch
Questions

Qwen3.8-Max FAQ

01What is Qwen3.8-Max?+
Qwen3.8-Max is Qwen's August 2026 flagship MoE model with 2.4 trillion total parameters and roughly 95 billion active parameters. It is designed for coding, research, professional work, and long-running agents.
02Does Qwen3.8-Max have a 1M context on SPIRITT?+
Not on the current SPIRITT route. The model developer documents a separate 1M multimodal service, while SPIRITT provides a 262K text-only context. This page reports both without blending them.
03Where does SPIRITT host Qwen3.8-Max?+
SPIRITT serves this Chinese-developed model through U.S.-operated infrastructure, independently of Qwen's own cloud. This describes the hosting operator and does not promise U.S.-only data residency.
04What is Qwen3.8-Max best for?+
Long repository work, multi-source research, professional deliverables, and tool-heavy tasks that benefit from sustained planning and iteration more than instant token streaming.
05What are the main limitations of Qwen3.8-Max?+
The SPIRITT route is text-only and capped at 262K context, Artificial Analysis measures roughly 21 output tokens per second, and the $2/$6 reference token price is higher than smaller value models. Vendor agent results also remain sensitive to the harness.
06What are the reference token economics for Qwen3.8-Max?+
The hosted route's reference pricing is $2 per million input tokens, $0.25 per million cached input tokens, and $6 per million output tokens. SPIRITT plan pricing and usage accounting are separate product terms.
07Can I connect a Qwen subscription from SPIRITT Subscriptions?+
Not currently. Choose Qwen3.8-Max through SPIRITT Models. The Subscriptions tab currently connects ChatGPT, Claude, and Grok accounts, not Qwen accounts.
08What is a good example of using Qwen3.8-Max well?+
Give it a repository migration or evidence-heavy market study: inspect the source, create a plan, implement or analyze in stages, run checks, and return both the deliverable and the proof that each acceptance condition passed.
09Where is Qwen3.8-Max best to test for full agentic capabilities?+
SPIRITT Workspaces. Pick Qwen3.8-Max in the model picker and run real work in a fully equipped cloud environment: tools, browser, files, terminal, memory, and durable sessions. That is where the model can act as an agent, not just chat.
Buy from builders who use what they sellBuilt usingSPIRITT