Qwen3.8 Max 0902 by NanoGPT | 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

As a dated refresh in Alibaba's Qwen3.8 Max line, this 0902 snapshot is positioned as a targeted improvement rather than a fresh pre-training cycle, concentrating gains on coding against larger codebases, long-horizon agent runs that can proceed without supervision, and vision handling that reads dense documents and charts more accurately. The dated identifier pins the snapshot in place so production traffic continues to hit the same revision, and gateway routing rules can transparently redirect earlier aliased requests onto the new release without any application-side change.

In practical terms, the model fits teams that need an assistant able to keep state across extended tasks, emit structured outputs that downstream tools can consume, and reason over mixed inputs including attached PDFs and video alongside text. Routing through aggregators exposes a direct Alibaba Cloud path with a relatively low cache-read cost, while gateway-level observability around latency and throughput helps operators gauge fit for live coding agents and document-heavy workflows before committing traffic at scale.

NanoGPTalibaba/qwen3.8-max-0902qwen

Open provider page Provider docs

Quick Info

Provider: NanoGPT
Model key: alibaba/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

Input tokens: 991,808 tokens
Output tokens: 131,072 tokens
Context window: 991,808 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

NanoGPT

Coverage

Alibaba's Qwen team has rolled out an updated build of its flagship Qwen3.8-Max line, designated Qwen3.8-Max-0902, debuting at the number 1 spot on Arena.ai's Code Arena WebDev leaderboard with a score of 1,691 points. According to Arena.ai's announcement, that result places the new build 3 points above Claude Opus 5 ( The Qwen3.8-Max-0902 model is reported as a 2.4 trillion parameter system and is priced at $2 per million input tokens, $6 per million output tokens, $0.17 per explicit cache hit, and $0.25 per implicit cache hit, equating to roughly a $5 blended cost per million tokens. Category-level Arena.ai rankings put it first in

NanoGPT

CoverageBenchmark

AI Release Tracker's model page for Qwen3.8-Max-0902 confirms the model was released by Qwen on Wednesday, September 2, 2026, arriving just 7 days after Qwen3.8-Flash-Next. The listed specifications are 2.4 trillion parameters and a 1 million token context window. The page catalogs a broad benchmark suite spanning NL2R On the tracked benchmarks, Qwen3.8-Max-0902 posts 64.9% on NL2Repo-Bench and 70% on QwenSWEBench V2, both listed as best published scores among tracked models, alongside 69.3% on DeepSWE 1.1 (5th of 24, behind Muse Spark 1.3's 75.4%) and 29% on Terminal-Bench 3.0, again a best published result. Long-horizon and workflo