What Is Qwen3.8-Max? A First Look at Alibaba's 2.4T-Parameter Flagship

Qwen3.8-Max is a 2.4-trillion-parameter multimodal large language model from Alibaba Cloud's Qwen team, announced on July 19, 2026 as the preview variant of the upcoming Qwen3.8 flagship. Alibaba claims it is "one of the most powerful models available today, compatible to leading frontier AI models, second only to Fable 5," and has released it as Qwen3.8-Max-Preview on the Alibaba Cloud Token Plan, Qoder, and QoderWork. The full Qwen3.8 model is scheduled to launch with open weights, though no date, license, or independent benchmarks have been published.

Key Takeaways

• Qwen3.8-Max has 2.4 trillion total parameters, per Alibaba's official launch announcement on July 19, 2026.
• The preview variant, Qwen3.8-Max-Preview, is live now on Alibaba Cloud Token Plan, Qoder, and QoderWork; the full model is not yet released.
• Alibaba positions the model as "second only to Fable 5" — an official self-assessment, unverified by third-party evals.
• Alibaba has committed to an open-weight release for Qwen3.8, but has published no date, license, or Hugging Face repository yet.
• Access starts at USD $6/month (39 CNY) via Token Plan Lite; there is no standalone per-token API price yet.
• A 1M-token context window is cited in Qwen Code v0.20.0 integration notes, but is not confirmed in an official model card.

What Is Qwen3.8-Max?

Qwen3.8-Max is the flagship tier of Alibaba's Qwen3.8 model family, the successor to Qwen3.7-Max. It is a multimodal model — the vendor and community both report image and video input alongside text — targeted at coding, agentic workflows, and long-horizon "professional cowork" tasks. It was announced simultaneously by the Alibaba Cloud account and the Qwen team, with the preview going live the same day.

Two names appear in early coverage and matter for anyone shopping the docs:

• Qwen3.8 — the full model, promised as an open-weight release "soon," with no date attached.
• Qwen3.8-Max-Preview — the currently accessible hosted variant, described by Alibaba as continuously evolving and subject to change or replacement when the production version ships.

The launch landed three days after Moonshot's Kimi K3 announcement, and Hacker News commenters read the timing as a direct response to Kimi K3's 2.8T open-weight positioning. Whatever the motivation, both drops signal that Chinese labs are competing at the trillion-parameter open-weight frontier on a weekly cadence.

Qwen3.8-Max at a Glance

Field	Value
Developer	Alibaba Cloud (Qwen team)
Type	Large language model, flagship tier
Total parameters	2.4 trillion
Architecture (dense vs MoE)	Not yet confirmed
Modality	Text, image input, video input (per vendor + community reports)
Context window	1M tokens (unconfirmed — cited in Qwen Code v0.20.0 notes)
Pricing (per-token API)	Not yet published
Token Plan pricing	From USD $6/month (Lite) / 39 CNY/month
Availability	Preview on Alibaba Cloud Token Plan, Qoder, QoderWork
License	Not yet announced; open-weight release promised
Release status	Preview live (July 19, 2026); full open-weight release TBD
Model ID	qwen3.8-max-preview

How Qwen3.8-Max Works and What Makes It Different

The public technical detail is thin. Alibaba's announcement highlights three claims and no architecture diagram:

1. 2.4T total parameters — vendor-reported, no active-parameter or MoE-vs-dense breakdown.
2. Continuously evolving preview — model behavior is expected to change during the preview window, and Alibaba reserves the right to swap or retire the endpoint.
3. Sharper code engineering and professional cowork — improvements over Qwen3.7-Max are framed around long-horizon coding and Office-style task execution rather than raw benchmark chasing.

Three coined terms recur in the launch materials and community coverage and are worth pinning down:

• Token Plan — Alibaba Cloud's subscription-and-credits access layer for Qwen3.8-Max-Preview, sold in Lite/Standard/Pro tiers instead of per-token metering.
• Qoder and QoderWork — Alibaba's agentic coding IDE and desktop assistant, respectively, where the model is offered with aggressive credit multipliers.
• Off-Peak Rate — Qoder's daily 14:00–00:00 UTC discount window where preview usage runs at a fraction of standard credits.

Qoder's own documentation lists Qwen3.8-Max-Preview at a 0.5x standard rate, dropping to 0.05x (90% off) during regular hours and 0.01x (98% off) off-peak, per its Premium Model Discount Rates page. The pricing is a promotional wrapper around a hosted preview — not a permanent list price.

What You Can Do With Qwen3.8-Max

Based on the launch materials and early hands-on posts, the intended use cases cluster around three areas:

• Full-stack coding and agentic development. The model plugs into Qwen Code, Claude Code, Cursor, Cline, OpenCode, and Codex-style clients through Token Plan's OpenAI/Anthropic-compatible endpoint. Early testers have run classic "clone this app" tasks, including a hands-on rebuild of a Mac cleanup utility posted within hours of launch.
• Long-horizon professional workflows. Alibaba specifically calls out Office-style automation and long-running agent tasks as areas where Qwen3.8-Max improves on 3.7-Max.
• Multimodal reasoning. Image and video inputs are supported per vendor description, though public examples so far focus on code, not vision.

For teams evaluating flagship models specifically for large-context agentic coding, our reference page on what Kimi K3 is covers the closest same-week comparison point.

How Qwen3.8-Max Compares

Alibaba's positioning statement — "second only to Fable 5" — sets up an implicit hierarchy that no third party has verified yet. The following comparison uses only vendor-reported numbers and community claims explicitly labeled as such.

Model	Total parameters	Open weights	Vendor claim	Status
Qwen3.8-Max	2.4T	Promised	"Second only to Fable 5"	Preview (Jul 19, 2026)
Kimi K3	2.8T	Yes (planned Jul 27)	Frontier-competitive	Announced
Claude Fable 5	Not disclosed	No	Frontier	Released

A widely shared take from Chubby's analysis captured the community mood: "Qwen-3.8-Max reportedly outperforms GPT-5.6 Sol and trails Fable 5 by only a narrow margin. I know this still needs to be verified." That verification hasn't landed.

Availability: How to Access Qwen3.8-Max

Qwen3.8-Max-Preview is available today through three official Alibaba surfaces:

1. Alibaba Cloud Token Plan (international) <!-- rel:dofollow class:authoritative --> — subscription plans starting at USD $6/month for Lite, $20/month for Standard, $70/month for Pro.
2. Token Plan (China) — 39 / 139 / 499 CNY tiers for the same Lite/Standard/Pro structure.
3. Qoder and QoderWork — Alibaba's agentic coding IDE and desktop assistant, where preview usage is heavily discounted during the launch window.

The Token Plan endpoint speaks both OpenAI and Anthropic protocols, which is why third-party clients including Claude Code, Cursor, Cline, and Codex work out of the box once an API key is issued. Alibaba has not yet published a standalone per-token price for a general-purpose API. If you want to try a comparable Qwen-family model that is already generally available with straightforward API pricing, Qwen Image 2.0 is a solid stand-in for image workloads while the language model preview stabilizes.

Alibaba Cloud describes the Token Plan Individual as "one plan, every modality — Qwen3.8-Max-Preview · GLM-5.2 · DeepSeek-V4-Pro." The multi-vendor bundling is unusual and suggests Alibaba is positioning Token Plan as a hub rather than a Qwen-only funnel.

What We Don't Know Yet

Several things a reference reader would reasonably expect are not yet public:

• Architecture details — dense vs MoE, active parameters per token, training-token count, training data cutoff.
• Independent benchmarks — no SWE-Bench, LMSYS Arena, GPQA, AIME, or agent-eval scores exist for Qwen3.8-Max as of publication.
• The open-weight release date and license — Alibaba has committed to open weights but not to a window or a license family.
• Whether smaller distillations will ship — early community requests focus on 27B and 35B variants for local use, but Alibaba has not confirmed any.
• Final GA pricing and rate limits — the preview may be replaced with a production model at different pricing.
• Confirmed context window — 1M is cited by community integrations, but Alibaba has not published an official spec sheet.

This page will be updated as those numbers land.

Frequently Asked Questions

Is Qwen3.8-Max open source?

Qwen3.8-Max is not yet open source, but Alibaba has stated the full Qwen3.8 release will go open-weight. Only the preview variant, Qwen3.8-Max-Preview, is currently accessible, and it runs as a hosted service on Alibaba platforms. A license and release date have not yet been confirmed.

How many parameters does Qwen3.8-Max have?

Qwen3.8-Max has 2.4 trillion total parameters, according to Alibaba's launch announcement. Alibaba has not published whether it is a dense or Mixture-of-Experts architecture, nor active-parameter counts. The 2.4T figure is a total-parameter claim from the vendor and has not been independently verified.

How much does Qwen3.8-Max cost?

Qwen3.8-Max-Preview is bundled inside Alibaba Cloud's Token Plan, which starts at USD $6/month (Lite) for international users and 39 CNY/month in China. Standalone per-token API pricing has not been officially released. On Qoder, preview usage runs at 0.05x credits during regular hours and 0.01x during off-peak, according to Qoder's event documentation.

Qwen3.8-Max vs Fable 5 — which is better?

Alibaba claims Qwen3.8-Max is second only to Anthropic's Fable 5 among frontier models, but no independent benchmarks support this yet. All performance rankings so far come from Alibaba's own launch text, not third-party evaluations. Community reactions treat the ranking as marketing until public evals land.

How do I access Qwen3.8-Max?

Qwen3.8-Max-Preview is accessible through Alibaba Cloud's Token Plan (international and China regions), Alibaba's Qoder agentic IDE, and QoderWork. Users subscribe to a Token Plan tier, receive an API key, and connect through supported clients such as Qwen Code, Claude Code, Cursor, or Cline. There is no standalone public API yet.

When will Qwen3.8-Max open weights be released?

Alibaba has said the open-weight release is coming soon but has not published a date. The phrasing in the launch tweet is "launching and going open-weight soon," with no confirmed window. Track the official Qwen account and Hugging Face for the actual drop.

What is the context window of Qwen3.8-Max?

Community reports and Qwen Code integration notes cite a 1M-token context window for Qwen3.8-Max-Preview, but Alibaba has not confirmed this in an official spec sheet. The number appears in the Qwen Code v0.20.0 release notes referenced by early testers. Treat 1M as unverified until an official model card publishes.

What to Watch Next

Three signals will determine whether Qwen3.8-Max lives up to its launch positioning. First, the actual open-weight drop — a repository on Hugging Face with a license attached, not just a promise. Second, independent benchmarks: SWE-Bench Verified, LMSYS Arena Elo, and any long-context agent eval that can pressure-test the claimed 1M window. Third, whether Alibaba ships smaller variants that make the 2.4T flagship story usable outside a hosted subscription.

Building similar large-context reasoning or agentic coding workflows? On kie.ai you can try Kimi K3, Claude Fable 5, and GPT-5.6.