Qwen 3.8 Max
2.4 trillion-parameter multimodal flagship model, claimed second only to Claude Fable 5
Context window
983616
Input / 1M tokens
2.00
Output / 1M tokens
6.00
Provider
Alibaba
Data verified 2026-08-05
Alibaba's flagship Qwen 3.8 Max is a 2.4 trillion-parameter sparse Mixture-of-Experts model and the team's first multimodal model above 1 trillion parameters. Currently in preview, it processes text, images, video, and documents. Alibaba claims improvements over Qwen 3.7-Max in coding, full-stack development, data analysis, and office workflows, though no public benchmarks have been published to verify the claim.
Capability index
Relative estimates (0-100) to place this model against its peers, grounded in published benchmarks.
How to access it
Available through Alibaba Token Plan subscriptions ($6-$68/month at 10% preview discount), Qoder coding platform, and QoderWork productivity tool. Credit-based token subscriptions; per-token API pricing not yet published.
Strengths
- ✓Large parameter count (2.4T total parameters)
- ✓Multimodal capabilities (text, image, video, documents)
- ✓API-compatible with both OpenAI and Anthropic protocols
- ✓Available through affordable subscription tiers during preview
- ✓Positioned for long-horizon agent and coding tasks
Best for developers who...
When to choose it (and when not to)
Reach for Qwen 3.8 Max when...
- →If seeking alternatives to Western frontier models
- →For cost-sensitive deployments during preview pricing
- →Multimodal workloads requiring visual understanding
Look elsewhere if...
- ✕Production workloads requiring stable APIs (preview status, continuously evolving)
- ✕Teams needing published benchmarks and verified performance claims
- ✕Use cases requiring open-weight or self-hosted deployment (not yet available)
- ✕Applications sensitive to model stability (preview may be replaced or updated)
How to use it
- ›Model supports reasoning controls (low, high, xhigh settings; xhigh is default)
- ›Compatible with existing OpenAI and Anthropic tooling without rewrites
- ›Multimodal input via text and images confirmed; video/document support reported but unconfirmed
Quickstart
Python# Access through Alibaba Token Plan during preview
# Standard per-token API pricing not yet published
# Currently available via credit-based subscriptions through qwencloud.comOfficial API quickstart code not available; model is in preview with documented context limit of ~984K tokens and 128K max output
API model id: qwen3.8-max
Benchmarks
| Benchmark | Score | Notes |
|---|---|---|
| Vendor Claim (Unverified) | Second only to Claude Fable 5 | No published benchmark table, model card, or independent testing available as of July 2026 |
| Terminal-Bench 2.1 | 86.6 | |
| SWE-bench Pro | 67.7 | |
| FrontierSWE | 73.5 | |
| PaperBench | 93.0 | |
| GPQA Diamond | 92.6 | |
| OSWorld-Verified | 86.1 |
Source: No public benchmarks published
Compare Qwen 3.8 Max
Compare Qwen 3.8 Max with any other model
Build a comparison →All model comparisons →