Alibaba’s Qwen3.8 Max lifts agentic-work scores to 53, with open weights planned next week
ArtificialAnlys · x · 2026-08-04
Alibaba’s Qwen3.8 Max shows clear gains on agentic work benchmarks, reaching 53 on the Artificial Analysis Intelligence Index and closing in on Claude Sonnet 5, while still trailing Kimi K3 on both score and cost.
- Scale & availability: Alibaba says it is a 2.4T-parameter MoE with 95B active parameters per forward pass, a 1M-token context window, and text/image/video input with text output.
- Benchmarks:
- Intelligence Index: 53, up from Qwen3.7 Max’s 46
- GDPval-AA: 1599 Elo, a 329-point jump over Qwen3.7 Max and effectively tied with Claude Sonnet 5
- AA-Briefcase: 1430 Elo, ahead of Claude Sonnet 5 and GLM-5.2, but behind Kimi K3
- Remaining gains are strongest in agentic coding; some science/knowledge metrics are flat or slightly down, including a larger hallucination rate on AA-Omniscience
- Cost: $1.76 per Intelligence Index task, roughly equal to Claude Sonnet 5, about 2× Kimi K3 and 3× GLM-5.2. Alibaba says the public weights release is planned for next week.
Related event: Alibaba releases Qwen3.8-Max, open-sources weights next week(44 posts)→
More from coding & agent
- Agent named 'Ilya' reportedly scores 1.1pt higher on benchmarks — daniel_mac8 · 2026-10-16
- Gemini CLI fix: substring matching inlined unrequested binary files into model context — priyanshu1976 · 2026-10-16
- Building AI agents on real healthcare claims data: the domain work that makes them work — alex_verem · 2026-10-16
- OpenAI stealth-testing mysterious internal model 'Aeon' in Codex desktop app and Dots — cedric_chee · 2026-10-16
- Vibe-coded iPhone Duo sniping bot built with Codex in 17 minutes, 41 tests passing — xiaohu · 2026-10-16
- Long-term multiplayer agent memory tool tops BEAM memory benchmark — CurieuxExplorer · 2026-10-16