China AI Hub
Models / Qwen3.8-Max
Qwen3.8-Max
Qwen3.8 family · Provider: alibaba-cloud · Status: active · Released: 2026-08
Qwen3.8-Max is a Qwen3.8 model developed by alibaba-cloud, a Chinese AI company, released 2026-08 with a 1,048,576-token context window and closed weights.
| Key fact | Value |
|---|---|
| Model ID | qwen3.8-max |
| Version | 0902 |
| Architecture | 2.4T-parameter MoE, 95B activated, 512 experts (10 routed + 1 shared per token), 92 layers, Gated DeltaNet + Gated Attention hybrid |
| Context window | 1,048,576 tokens |
| Max output | 131,072 tokens |
| Open weights | No |
| License | proprietary |
| Self-hosting | No |
| API available | Yes |
| API pricing | $2 input / $6 output per 1M tokens (USD) · provider pricing page |
| Regions | china-beijing, singapore, hong-kong, germany-frankfurt, us-virginia, japan-tokyo |
| Cloud providers | Alibaba Cloud |
Capabilities
| Capability | Supported |
|---|---|
| Reasoning | Yes |
| Coding | Yes |
| Math | Unknown |
| Chinese | Unknown |
| English | Unknown |
| Multilingual | Unknown |
| Vision | Yes |
| Audio | Unknown |
| Video | Yes |
| Tool calling | Yes |
| Function calling | Unknown |
| Structured output | Yes |
| Agent capability | Unknown |
| RAG | Unknown |
| Computer use | Unknown |
Benchmark results
| Benchmark | Version | Score | Metric | Date | Source type | Source |
|---|---|---|---|---|---|---|
| Terminal-Bench 2.1 | — | 86.6 | accuracy | 2026-08 | vendor_reported | link |
| SWE-bench Pro | — | 67.7 | accuracy | 2026-08 | vendor_reported | link |
| GPQA Diamond | — | 92.6 | accuracy | 2026-08 | vendor_reported | link |
| HLE | — | 43.6 (56.2 with tools) | accuracy | 2026-08 | vendor_reported | link |
| MRCR v2 256K | — | 92.9 | accuracy | 2026-08 | vendor_reported | link |
Benchmark scores are single data points, not universal rankings. Vendor-reported scores are labeled as such.
Known limitations
- Closed API model (the open Qwen3.8-2.4T-A95B weights are text-only and thinking-only - not the same product)
- Exact API release date not stated; the 0902 snapshot is dated 2026-09-02
- Prices differ by region: Singapore $2/$6; Beijing and Global regions $1.65/$4.951 per 1M tokens
- Benchmarks are from the vendor model card (Qwen3.8-Max column); not independently verified
Qwen3.8-Max is Alibaba’s flagship API model: a 2.4T-parameter MoE (95B activated) with a 1M-token context window (991,808 max input), 131,072 max output and 262,144 max chain-of-thought. Input modalities are image, text and video; output is text. It supports thinking and non-thinking modes, function calling, structured outputs, web search, prefix completion, context caching and batch inference (Beijing only).
The 0902 snapshot (2026-09-02) upgraded coding and vision. API pricing: Singapore $2 input / $6 output per 1M tokens; Beijing and other Global regions $1.65 / $4.951 (as of 2026-09-20). Fine-tuning is not supported.
Provider
Pricing
Benchmarks with results for this model
Agents built on this model
Related models (same family)
API
Model family timeline
| Model | Released | Status |
|---|---|---|
| Qwen3.8-Max (this page) | 2026-08 | active |
| Qwen3.8-Flash | 2026-08 | active |
| Qwen3.8-2.4T-A95B | 2026-08-12 | active |
Data interpretation
| Field | Value | Evidence type |
|---|---|---|
| Context window | 1,048,576 tokens | Official |
| Architecture | 2.4T-parameter MoE, 95B activated, 512 experts (10 routed + 1 shared per token), 92 layers, Gated DeltaNet + Gated Attention hybrid | Official |
| Open weights | No | Official |
| License | proprietary | Official |
| API pricing | $2 / $6 per 1M tokens (USD) | Official |
| Terminal-Bench 2.1 | 86.6 accuracy | Vendor-reported |
| SWE-bench Pro | 67.7 accuracy | Vendor-reported |
| GPQA Diamond | 92.6 accuracy | Vendor-reported |
| HLE | 43.6 (56.2 with tools) accuracy | Vendor-reported |
| MRCR v2 256K | 92.9 accuracy | Vendor-reported |
| Family position | 3 models in the Qwen3.8 family | China AI Hub analysis |
Evidence types: Official = vendor documentation, pricing pages or model cards. Vendor-reported = benchmark scores published by the vendor. China AI Hub analysis = derived from the database itself. See the sourcing policy.
Sources
Confidence and source hierarchy per the sourcing policy. Facts change; check the source before relying on this page.
What is the context window of Qwen3.8-Max?
Qwen3.8-Max has a 1,048,576-token context window and a maximum output of 131,072 tokens.
Is Qwen3.8-Max open weight?
No — Qwen3.8-Max is not open weight.
How much does Qwen3.8-Max cost through the API?
$2 per 1M input tokens and $6 per 1M output tokens (USD).
Where does China AI Hub get its Qwen3.8-Max data?
From 3 sources (official pages first), last verified 2026-09-20. Benchmark scores are labeled by source type; see the sourcing policy for details.