Epoch Capabilities Index (ECI) ↗142.33 ECI
Qwen3-Max · General ECI · 26 October 2026 fit ↗
Qwen / MODEL EVIDENCE
Qwen3 Max
16 benchmarks · 20 observations · 16 sources · released 2025-09-24
Scores at a glance
16 benchmarks · large dot: Qwen3 Max; small dots: closest models where reported
LegalBench ↗81.86 %
Qwen 3 Max · LegalBench · Accuracy ↗
CaseLaw v2 ↗47.48 %
Qwen 3 Max · CaseLaw · Accuracy ↗
1 more reported result
54.98 % · Qwen 3 Max Thinking · CaseLaw · Accuracy · Vals AI · CaseLaw ↗
CorpFin v2 ↗55.94 %
Qwen 3 Max · CorpFin · Accuracy ↗
1 more reported result
68.03 % · Qwen 3 Max Thinking · CorpFin · Accuracy · Vals AI · CorpFin ↗
MortgageTax & TaxEval v2 ↗73.51 %
Qwen 3 Max · TaxEval · Accuracy ↗
MedCode ↗31.37 %
Qwen 3 Max Thinking · MedCode · Accuracy ↗
MedScribe ↗72.71 %
Qwen 3 Max Thinking · MedScribe · Accuracy ↗
MedQA (USMLE) ↗87.37 %
Qwen 3 Max · MedQA · Accuracy ↗
AIME 2024–2025 · Vals ↗81.04 %
Qwen 3 Max · AIME · Accuracy ↗
MGSM (Multilingual GSM) ↗91.82 %
Qwen 3 Max · MGSM · Accuracy ↗
GPQA Diamond ↗79.55 %
Qwen 3 Max · GPQA Diamond · Accuracy ↗
1 more reported result
84.85 % · Qwen 3 Max Thinking · GPQA Diamond · Accuracy · Vals AI · GPQA Diamond ↗
MMLU-Pro ↗84.36 %
Qwen 3 Max · MMLU Pro · Accuracy ↗
1 more reported result
84.98 % · Qwen 3 Max Thinking · MMLU Pro · Accuracy · Vals AI · MMLU Pro ↗
Vibe Code Bench ↗3.51 %
Qwen 3 Max (OpenHands) · Vibe Code Bench · original task set ↗
LiveCodeBench ↗78.22 %
Qwen 3 Max · LiveCodeBench · Accuracy ↗
ALE-Bench ↗370.45 Performance points
qwen3-max · Average contest performance ↗
Vending-Bench 2 ↗71.565 USD
Qwen3 Max · End-of-year balance ↗
Closest models by score
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
6 of 9 shared benchmark comparisons
6 of 9 shared benchmark comparisons
6 of 9 shared benchmark comparisons
How similarity is measured
Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.
Sources
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
- Ale Bench — methodology and published results ↗
- Vending Bench 2 — methodology and published results ↗
- AI benchmarks & capabilities ↗
- AIME — evaluation methodology and results ↗
- CaseLaw — evaluation methodology and results ↗
- CorpFin — evaluation methodology and results ↗
- GPQA Diamond — evaluation methodology and results ↗
- LiveCodeBench — evaluation methodology and results ↗
- LegalBench — evaluation methodology and results ↗
- MedCode — evaluation methodology and results ↗
- MedQA — evaluation methodology and results ↗
- MedScribe — evaluation methodology and results ↗
- MGSM — evaluation methodology and results ↗
- MMLU Pro — evaluation methodology and results ↗
- TaxEval — evaluation methodology and results ↗
- Vibe Code Bench — evaluation methodology and results ↗
Names used in source tables
Qwen 3 Max · Qwen 3 Max (OpenHands) · Qwen 3 Max Thinking · Qwen3 Max · Qwen3-Max · qwen3-max