Alibaba Releases 2.4T Parameter Qwen3.8-Max, Open-Sourcing Next Week
千问大模型 · wechat · 2026-08-03
Alibaba's Qwen has officially released its most powerful model to date, Qwen3.8-Max, with 2.4 trillion parameters (95B active) and a 1 million token context window. Model weights for both Max and 27B versions will be open-sourced next week.
Key Capabilities & Benchmarks:
- Autonomous Coding & Research: The model can independently build self-evolving codebases over 16 days. Given just a research paper and GPUs, it worked autonomously for 5 days to reproduce the study and even proposed a new method that surpassed the original paper's results.
- Long-horizon Tasks & Multimodal Agents: It autonomously optimized a hardware accelerator design from 8,298 logic gates down to 678. It also won a simulated e-commerce management benchmark with a 4.16x return on investment. The hybrid agent capability combines coding with GUI operation to replicate real-world apps from scratch.
- Dynamic Workflows: In a single conversation, it orchestrated 330 sub-agents to run 6,000 backtests, delivering a profitable quantitative trading strategy end-to-end.
API Pricing: Priced at $2/M input and $6/M output tokens internationally, offering high cost-effectiveness.
Related event: Alibaba releases Qwen3.8-Max, open-sources weights next week(44 posts)→
More from coding & agent
- DeepMind turns coding agent Antigravity into a science tool, surfacing novel antimicrobial peptide in one session — pushmeet · 2026-10-16
- Cross-review coding workflow: Claude writes the plan, Codex audits it — theteknosaur · 2026-10-16
- Simon Willison: code mode is the future of tool calling, if it can be sandboxed — springrod · 2026-10-16
- Normalizing Claude's schema-free MCP output before it hits downstream n8n nodes — snowingbol · 2026-10-16
- Agent named 'Ilya' reportedly scores 1.1pt higher on benchmarks — daniel_mac8 · 2026-10-16
- A plain notes.md fixed my agent's cross-session amnesia better than chat history — rahul_watertech · 2026-10-16