Qwen3.8 Max 0902 by OpenRouter | AI model information

$10 off the fastest Kimi K3 and GLM 5.3$10 off the fastest Kimi K3 and GLM 5.3

Model details

Qwen3.8 Max 0902

Qwen3.8 Max 0902 is a dated, post-trained snapshot from Alibaba’s Qwen team, retaining the family’s sparse mixture-of-experts architecture with 2.4 trillion total parameters and 95 billion active parameters. Its training emphasis shifted toward coding and “cowork” tasks, including complex engineering projects, multi-tool orchestration, and longer autonomous workflows, while also targeting chart reasoning, document parsing, and multimodal analysis.

The snapshot is best suited to repository comprehension, structured professional work, and tool-assisted tasks where its broad context and multimodal understanding can be applied over extended materials. Published comparisons show substantial gains over the earlier snapshot on terminal coding, software replication, and job-task evaluations, but the results are not uniformly dominant: competing systems still lead several long-horizon coding and office-work measures. Its strongest practical fit is therefore as a versatile agentic model for iterative coding and document workflows, with workload-specific evaluation still important.

OpenRouterqwen/qwen3.8-max-0902qwen

Open provider page Provider docs

Quick Info

Provider: OpenRouter
Model key: qwen/qwen3.8-max-0902
Release date: Sep 2, 2026
Last updated: Sep 2, 2026
Input modalities
Output modalities
Capabilities

Cost

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

Limits

Output tokens: 131,072 tokens
Context window: 1,000,000 tokens

Transparent token rates

Compare Qwen3.8 Max 0902 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 0902

OpenRouter

Official sourceBenchmark

The canonical /qwen3.8-max-0902 model page restates the core spec: 2.4T-parameter MoE, 1M-token context, multimodal text/image/video input, reasoning by default, tool calling and structured outputs, released Sep 3, 2026. Alibaba Cloud Int. is the only provider, with the same $2/$6 price points and $0.25 cache reads. It The page provides a price-history chart and token-share data useful for budgeting, and the uptime/availability figures reinforce reliability for production traffic. The single-provider dependency remains a constraint, and the rolling 3-day window means these metrics will shift; nonetheless this is the strongest first-p

OpenRouter

Official sourceBenchmark

The dedicated OpenRouter API page for Qwen3.8 Max (0902) confirms Alibaba Cloud International as the sole host provider, with listed pricing at $2.00/M input and $6.00/M output, $0.25/M cache reads, and $2.50/M cache creation. Operational metrics show 1.15s P50 latency, 40 tok/s throughput, 100% uptime and 99.87% avail Despite its older Aug 3, 2026 publishedAt timestamp, the page content clearly describes the 0902 snapshot, suggesting a URL canonicalization or metadata staleness issue. Single-provider hosting means there is no failover diversity on OpenRouter for this model, and the 40 tok/s throughput is modest for a 2.4T-parameter

OpenRouter

CoverageBenchmark

DataCamp reports that Alibaba released Qwen3.8-Max-0902 on September 2, 2026 as a post-training upgrade focused on coding and long-horizon agent work, with architecture and pricing unchanged from the base model ($2/M input, $6/M output, 1M context). The headline result is a first-place finish on Code Arena WebDev with Against Claude Opus 5, the 0902 snapshot leads on three coding benchmarks (MLS-Bench-Lite, SWE-Atlas QnA, QwenSWEBench V2), WorkArena Elo, and both multimodal evals (MMMU-Pro 82.7, ERQA 78.3), while Claude still leads TerminalBench 3.0 (42.7), DeepSWE 1.1 (73.6), and several agent coordination tests. DataCamp cites Ali