How to Access the Qwen3.8 Max API
Short version: there is no public API yet. Here are the three paths that actually work today — and what to do if you need production access.
As of 2026-07-21, qwen3.8-max-preview is Token Plan-exclusive. It is not on OpenRouter, has no pay-as-you-go API, and no announced production pricing. The three working access paths:
- Qwen Cloud Token Plan — subscription from $6/month (limited-time), used inside compatible coding/agent tools
- Qoder — Alibaba's coding platform
- QoderWork — the agent workspace variant
Path 1: Token Plan (most flexible today)
Token Plan is a credit-based subscription. During the preview window, qwen3.8-max-preview invocations consume credits at as low as 10% of standard rates (official docs, 2026-07-21) — Alibaba is effectively subsidizing trial usage.
| Personal tier | Price (limited-time) | Credits |
|---|---|---|
| Lite | $8 → $6/month | 700 credits / 5-hour window |
| Standard | $25 → $18/month | 3,000 credits / 5-hour window |
| Pro | $80 → $68/month | 12,000 credits / 5-hour window |
Team tiers ($20–$200/seat/month) and full pricing details: see the pricing page. Several news outlets report $20/$70 for personal tiers — the official docs say $18/$68.
The flow, per official documentation: subscribe → create a plan API key → select the exact model ID qwen3.8-max-preview inside a compatible tool (Qwen Cloud documents integrations with coding agents and IDE tools). Use the full model ID — there are similarly-named older models (Qwen3-8B) that are unrelated.
Path 2 & 3: Qoder and QoderWork
Alibaba's own coding platform and agent workspace both expose qwen3.8-max-preview directly. If you just want to evaluate the model's coding behavior without configuring anything, this is the fastest route.
What about OpenRouter / third-party APIs?
Not available. People are searching for "qwen 3.8 max openrouter" — we checked: as of 2026-07-21 no third-party router carries it, because Alibaba has not opened any API beyond Token Plan. Any site claiming to sell "Qwen3.8 Max API" access today is reselling Token Plan, proxying Qoder, or misrepresenting an older model.
Need production Qwen access right now?
Until a real API ships, the practical setup teams use: run generally-available Qwen models (such as qwen3.7-max — same 1M context window, standard availability) in production, and evaluate qwen3.8-max-preview through Token Plan in parallel.
Call Qwen models via EvoLink's unified API →
EvoLink is a pay-as-you-go gateway for 50+ models including the Qwen family; we plan to route Qwen3.8 Max the moment a production API opens. (Disclosure: this site is run by the EvoLink team — about.)
FAQ
When will the public API launch?
Not announced. Alibaba has committed to open weights (no date) and the preview naming implies a GA release, but no timeline exists. We update this page when that changes.
Is there a free way to try it?
No standalone free tier as of 2026-07-21. The $6 Lite Token Plan month is currently the cheapest entry, with preview usage at ~10% credit consumption.
Can the preview model change under me?
Yes — that's what preview means. Alibaba can revise or replace it, and Preview models can drift between dates. Pin your evaluations to dates when comparing results.
Verified against official Qwen Cloud documentation on 2026-07-21. We deliberately publish no API code samples yet: no public endpoint exists, and we only publish code we have actually run.
Sources: Token Plan overview · Qwen Cloud model docs · OpenRouter Qwen listing (checked 2026-07-21)