Alibaba · source page ↗ · last checked Aug 22, 2026, 12:04 PM

Reported benchmarks · Qwen3.8-Max

captured Aug 22, 2026, 12:04 PM

Benchmark	Score
GPQA Diamond	92.6% · Vendor-run; independent replication 92.6% vs claimed 92.7%
SWE-bench Pro	67.7% · Alibaba's own corrected task set with Claude Code harness
Terminal-Bench 2.1	86.6% · 5-hour timeout; official leaderboard caps at 83.8%
PaperBench	93.0% · 28-point jump from Qwen3.7-Max; reproduces research paper results
IFBench	82.8% · Instruction following benchmark
OSWorld-Verified	86.1% · Desktop OS interaction; computer-use agentic workflows
Humanity's Last Exam	43.6% · Weakest showing among four flagship comparisons
MathVision	95.2% · Multimodal benchmark; vision-based math tasks

Vendor-reported via automated web search — not independently verified. See the cited matrix on /models.

Change history

• Vendor claim Aug 22, 2026, 12:04 PM

  Alibaba reported benchmarks updated

  Qwen3.8-Max: 8 benchmark claims (via web search)

• Vendor claim Aug 15, 2026, 6:03 AM

  Alibaba reported benchmarks updated

  Qwen3.8-Max: 10 benchmark claims (via web search)

• Vendor claim Aug 8, 2026, 12:03 AM

  Alibaba reported benchmarks updated

  Qwen3.8-Max: 8 benchmark claims (via web search)

• Vendor claim Jul 31, 2026, 6:03 PM

  Alibaba reported benchmarks updated

  Qwen3-235B-A22B: 9 benchmark claims (via web search)

• Vendor claim Jul 24, 2026, 12:03 PM

  Alibaba reported benchmarks updated

  Qwen3.7-Max: 10 benchmark claims (via web search)

• Vendor claim Jul 17, 2026, 6:03 AM

  Alibaba reported benchmarks updated

  Qwen3.7-Max: 10 benchmark claims (via web search)

• Vendor claim Jul 10, 2026, 12:03 AM

  Alibaba reported benchmarks updated

  Qwen3.7-Max: 12 benchmark claims (via web search)

• Vendor claim Jul 2, 2026, 6:03 PM

  Alibaba reported benchmarks updated

  Qwen3.7-Max: 12 benchmark claims (via web search)

• Vendor claim Jun 25, 2026, 4:22 PM

  Alibaba reported benchmarks updated

  Qwen3.7-Max: 11 benchmark claims (via web search)