Qwen3.8 Max Preview

Qwen3.8 Max Preview is a 2.4-trillion-parameter multimodal large language model from Alibaba Cloud's Qwen team, announced on July 19, 2026 as a preview variant of the upcoming Qwen3.8 flagship. Alibaba positioned the model as "second only to Fable 5," framing it as a frontier-tier system, though that self-assessment has not been independently benchmarked. The preview build is being distributed through Alibaba Cloud surfaces, with an official blog post titled "Qwen3.8-Max: A New Bar for Coding and Cowork" amplifying the launch on qwen.ai and drawing substantial Hacker News discussion.

The preview is being promoted as a step forward for coding assistance and agent-style cowork workflows, carrying both text and image inputs and producing text outputs suitable for long-running tasks. Reports indicate Alibaba has committed to releasing Qwen3.8 with open weights, and a Qwen3.8-27B local variant has been slated for an open-weight release shortly after the preview window. For teams evaluating general-purpose assistants that need strong reasoning, tool calling, and a very large context window, Qwen3.8 Max Preview fits as a cutting-edge option, while practitioners wanting fully open, locally deployable weights should wait for the planned open releases.

Charm Hyperqwen3.8-maxqwen

Open provider page Provider docs

Quick Info

Provider: Charm Hyper
Model key: qwen3.8-max
Release date: Jul 19, 2026
Last updated: Jul 19, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $2.00
Output token cost: $6.00

Limits

Output tokens: 65,536 tokens
Context window: 1,000,000 tokens

Transparent token rates

Compare Qwen3.8 Max Preview pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Qwen3.8 Max Preview

Charm Hyper

CoveragePreview

According to the BigGo Finance report, Alibaba Cloud released the Qwen3.8-Max preview on July 19, 2026, framing it as the company's first native multimodal model to cross the trillion-parameter threshold at 2.4 trillion parameters on a Mixture-of-Experts (MoE) architecture. The piece states Alibaba committed to returni The same article reports benchmark gains versus the prior generation of +22.8% on coding and +44.4% on office productivity scenarios, and mentions a "90% limited-time discount" and "daily update model" feedback program, drawing polarized reactions. The technical figures are presented without named benchmark suites or i

Charm Hyper

CoveragePreview

A third-party status page (qwen3coder.com) consolidates verified facts about the Qwen3.8-Max Preview model as of July 21, 2026. The exact hosted model ID is qwen3.8-max-preview, listed in Qwen Cloud's Token Plan model table with a 1M-token context window, and documented support for thinking, function calling, and built The page records that Qwen announced a new preview build on July 20, 2026, with claimed broad gains including a substantial improvement on web frontend work, and that the preview is updated daily as feedback arrives. The page explicitly flags these performance claims as vendor-reported without a public benchmark table,