Know when the route is verified
Receive an update when EvoLink confirms availability, the request model ID, pricing, and the API behaviors that have passed validation.
Qwen3.8 Max Preview API
Qwen / Alibaba-Text and vision understanding-Verified July 21, 2026 · Official preview · EvoLink route pending
Qwen3.8 Max Preview is the 2026 versioned preview—not the older Qwen3-8B model. It is available through Qwen Token Plan for interactive evaluation, while an EvoLink production API route and pricing remain pending.
Official status
Preview available through Qwen Token Plan
Official preview model ID
qwen3.8-max-preview
EvoLink API pricing
Not published
EvoLink Early Access
Get Qwen3.8 Max updates on EvoLink
Be among the first to know when EvoLink confirms a Qwen3.8 route, request model ID, pricing, and supported API behavior.
We will only email you about Qwen3.8 availability. No spam.
Want faster updates? Join Discord for the latest news
Why track Qwen3.8 through EvoLink?
Separate official preview access from a production API route, then compare the model without rebuilding your integration around an unstable identifier.
Keep one integration surface
Prepare to evaluate Qwen3.8 alongside supported Qwen, Kimi, Claude, GPT, and other models through one EvoLink account and gateway.
Compare before migrating
Replay your own coding, agent, visual-understanding, and long-context tasks before assigning production traffic to a preview model.
Qwen3.8 Max Preview features confirmed so far
These facts come from Qwen's official Token Plan and API documentation. They describe a channel-specific preview, not a final GA model or an EvoLink route.
Preview model identity
Qwen lists qwen3.8-max-preview as the exact model ID in its Token Plan documentation and says the preview may continue evolving before it is removed or replaced by a final version.
Reasoning and text generation
Qwen classifies the preview as a reasoning and text-generation model. Its OpenAI-compatible Chat API documentation includes low, medium, and xhigh reasoning-effort controls for this preview.
Visual understanding
The official Token Plan model list includes visual understanding for qwen3.8-max-preview. Endpoint-specific media limits still need verification before production use.
1M context and three cache modes
Qwen's current model list documents a 1M context window, while its cache guide lists explicit, implicit, and Responses API session caching for the Token Plan preview. These are Qwen-channel rules, not yet EvoLink route terms.
Built-in harness tools
Qwen documents web search, code interpreter, web extraction, reverse-image search, and text-based image search for the Token Plan preview channel.
Evaluation-only personal-plan access
Qwen limits the personal Token Plan to interactive coding and agent tools. Its subscription key cannot be used for automated scripts, custom application backends, or non-interactive batch calls.
Where teams should evaluate Qwen3.8 first
These are evaluation priorities based on the preview's documented positioning, not claims that Qwen3.8 already outperforms production alternatives.
Repository-level coding
Test multi-file changes, dependency tracing, implementation quality, and whether the result survives lint, build, and automated tests.
Long-running agent tasks
Measure tool selection, recovery after failures, instruction retention, looping behavior, and reviewer intervention over extended sessions.
Visual and document understanding
Evaluate screenshots, diagrams, dense documents, and mixed text-image inputs against task-specific acceptance criteria.
Research and productivity workflows
Test evidence synthesis, data analysis, report generation, and office workflows while separating model reasoning from built-in tool output.
What must be verified before production routing
A preview listing is not enough for a production migration. EvoLink will publish route-specific facts only after they are confirmed and tested.
Final model ID and lifecycle
Will the preview ID remain valid, redirect to a stable alias, or be replaced when Qwen3.8 reaches a broader release?
EvoLink availability and pricing
The EvoLink request ID, input/output pricing, cache treatment, and billing behavior have not yet been published.
Rate limits and regional access
Production teams need verified concurrency, RPM/TPM, region, quota, and availability behavior for the actual route they will use.
Tool and structured-output reliability
Function arguments, schema adherence, tool recovery, streaming, and multi-turn reasoning history need route-level testing.
Latency and successful-task cost
Preview credit promotions are not per-token API prices. Measure end-to-end latency, retries, output length, and human correction time.
Open-weight release details
Qwen has announced an open-weight direction, but the final repository, license, architecture details, and release timing still require confirmation.
What makes Qwen3.8 Max Preview distinctive?
Qwen documents a 1M-token context window together with visual understanding, extended reasoning controls, three cache modes, and Token Plan tools. That combination makes the preview relevant to repository-scale coding and long-running agents—but Token Plan access is not the same as a production backend API route.
Prepare for Qwen3.8 without pausing development
Follow the verified release facts, compare the closest alternatives, and prepare a replayable production evaluation.
Qwen3.8 features and release status
See what Qwen has officially documented, what remains unconfirmed, and what developers should prepare.
Qwen3.8 vs Kimi K3
Compare coding, agents, context, openness, access maturity, and production evaluation priorities.
Qwen3.8 benchmark evidence
Separate official positioning from independent evidence and build a workload-based benchmark plan.
Qwen3.8 vs Qwen3.7 Max
Decide what to test before replacing a documented production route with a changing preview.
Buy Qwen Token Plan
Use Qwen's official subscription page for interactive preview access. Token Plan is not an application-backend API plan.
Start with Kimi K3
Use an available model to establish coding, agent, and multimodal acceptance tests before Qwen3.8 activation.
Qwen3.8 Max API - FAQ
Is Qwen3.8 Max officially released?
Qwen3.8 Max Preview is officially listed through Qwen Token Plan as of July 21, 2026. It remains a preview that may change, be removed, or be replaced by a final version.
Can I use Qwen3.8 Max on EvoLink now?
Not yet. EvoLink has not published a supported Qwen3.8 route. Join early access for confirmed availability, model ID, pricing, and request behavior.
What is the official preview model ID?
Qwen's Token Plan documentation lists qwen3.8-max-preview. This is a channel-specific preview ID and should not be assumed to be the final EvoLink request ID.
Is Qwen3.8 the same as Qwen3-8B?
No. Qwen3.8 is a 2026 model-family version number, while Qwen3-8B is an older 8-billion-parameter model in the Qwen3 family. Downloads, local-deployment guides, and prices for Qwen3-8B do not describe Qwen3.8 Max Preview.
What is the difference between Qwen Token Plan and a standard API?
Token Plan is a subscription channel for interactive coding and agent tools with a dedicated key and usage credits. Qwen prohibits using that subscription key for automated scripts or application backends, so it is not a substitute for a standard production API route.
Does Qwen3.8 support prompt caching?
Qwen's current cache documentation lists explicit, implicit, and Responses API session caching for qwen3.8-max-preview. Creation cost, hit discounts, minimum lengths, and retention differ by mode; EvoLink-specific cache behavior remains pending until a route is activated.
Can I use a personal Token Plan key in my application backend?
No. Qwen limits the personal Token Plan to interactive coding and agent tools and prohibits automated scripts, custom application backends, and non-interactive batch use. Wait for an approved production route.
How much will Qwen3.8 cost on EvoLink?
EvoLink pricing has not been published. Token Plan subscription Credits and promotional multipliers should not be converted into an assumed per-token EvoLink price.
Does Qwen3.8 support images?
Qwen's official Token Plan model list marks qwen3.8-max-preview for visual understanding. Exact input formats and route limits still need verification for each production channel.
Is Qwen3.8 open weight?
Qwen has announced that open weights are planned, but a final repository, license, architecture disclosure, and release date were not confirmed as of July 21, 2026.
Should I replace Qwen3.7 Max now?
Not for production by default. Keep a documented stable route while you replay representative tasks on the preview and verify quality, latency, reliability, and successful-task cost.
Where can I buy Qwen3.8 Token Plan access?
Buy Token Plan only from Qwen's official pricing page at qwencloud.com/pricing/token-plan. It is intended for compatible interactive coding and agent tools, not automated application backends.
Models available on EvoLink today
Kimi K3
Available for coding, agent, long-context, and multimodal workloads with a documented EvoLink route.
View model
Claude Opus 4.8
A high-capability route for complex coding, long-running agents, research, and production review.
View model
GPT-5.6
A production reasoning family with Sol, Terra, and Luna tiers for different quality and cost targets.
View model