Alibaba Cloud / Qwenopen-weight
Released: 2026-09-23Qwen 3.8 Max Prime
Alibaba premier frontier open-weight reasoning flagship with dual deliberate thinking modes.
Intelligence Index
93.0/ 100
Calibrated multi-domain compositeThroughput Speed
68tok/sec
Output streaming throughputTime To First Token
390ms
Initial chunk server latencyLiveBench Contam-Free
80.0%
September 2026 suite scoreSourced Benchmark Evaluations (61)
Verified performance across authoritative benchmarks with provenance tracking.| Benchmark | Status | Score | Trust | Date | Source Type | Provenance |
|---|---|---|---|---|---|---|
| RULER | Nearing Saturation | 85.6% | 84 | 2026-09-12 | independent | Source ↗ |
| Mercury | Deprecated | 74.8% | 10 | 2026-09-12 | independent | Source ↗ |
| MGSM | Nearing Saturation | 88.4% | 65 | 2026-09-12 | independent | Source ↗ |
| MMLU | Saturated | 94.9% | 25 | 2026-09-12 | independent | Source ↗ |
| MMLU-Pro | Nearing Saturation | 85.6% | 75 | 2026-09-12 | independent | Source ↗ |
| MMLU-Redux | Saturated | 91.1% | 25 | 2026-09-12 | independent | Source ↗ |
| MT-Bench | Nearing Saturation | 8.8% | 75 | 2026-09-12 | independent | Source ↗ |
| MultiPL-E | Deprecated | 78.2% | 10 | 2026-09-12 | independent | Source ↗ |
| Needle-in-a-Haystack | Saturated | 98.6% | 44 | 2026-09-12 | independent | Source ↗ |
| OlympiadBench | Active | 76.3% | 80 | 2026-09-12 | independent | Source ↗ |
| Omni-MATH | Deprecated | 68.8% | 10 | 2026-09-12 | independent | Source ↗ |
| OpenAI-MRCR | Active | 78.1% | 80 | 2026-09-12 | independent | Source ↗ |
| PutnamBench | Active | 48.4% | 90 | 2026-09-12 | independent | Source ↗ |
| QASPER | Nearing Saturation | 77.2% | 65 | 2026-09-12 | independent | Source ↗ |
| QuALITY | Nearing Saturation | 81.8% | 78 | 2026-09-12 | independent | Source ↗ |
| AGIEval | Saturated | 80% | 25 | 2026-09-12 | independent | Source ↗ |
| SCROLLS | Nearing Saturation | 74.4% | 65 | 2026-09-12 | independent | Source ↗ |
| SimpleQA | Active | 48.4% | 90 | 2026-09-12 | independent | Source ↗ |
| Soohak | Active | 70.7% | 98 | 2026-09-12 | independent | Source ↗ |
| StructFlowBench | Nearing Saturation | 72.7% | 75 | 2026-09-12 | independent | Source ↗ |
| SuperGLUE | Saturated | 94.9% | 25 | 2026-09-12 | independent | Source ↗ |
| SWE-bench | Deprecated | 57.8% | 10 | 2026-09-12 | independent | Source ↗ |
| SWE-bench Pro | Active | 52.7% | 90 | 2026-09-12 | independent | Source ↗ |
| SWE-bench Verified | Saturated | 66.3% | 25 | 2026-09-12 | independent | Source ↗ |
| SWE-Lancer | Deprecated | 57.8% | 20 | 2026-09-12 | independent | Source ↗ |
| TabMWP | Saturated | 85.6% | 25 | 2026-09-12 | independent | Source ↗ |
| Terminal-Bench | Active | 59.5% | 80 | 2026-09-12 | independent | Source ↗ |
| TruthfulQA | Saturated | 81.8% | 25 | 2026-09-12 | independent | Source ↗ |
| WinoGrande | Saturated | 96.7% | 28 | 2026-09-12 | independent | Source ↗ |
| ZeroSCROLLS | Nearing Saturation | 74.4% | 78 | 2026-09-12 | independent | Source ↗ |
| MBPP | Saturated | 90.1% | 25 | 2026-09-12 | independent | Source ↗ |
| AgentIF | Nearing Saturation | 58.2% | 80 | 2026-09-12 | independent | Source ↗ |
| AIME | Nearing Saturation | 88.4% | 75 | 2026-09-12 | independent | Source ↗ |
| ARC-AGI | Saturated | 87.4% | 25 | 2026-09-12 | independent | Source ↗ |
| ARC-AGI-2 | Nearing Saturation | 72.5% | 75 | 2026-09-12 | independent | Source ↗ |
| ARC-AGI-3 | Active | 60.5% | 93 | 2026-09-12 | independent | Source ↗ |
| BIG-Bench Hard | Saturated | 91.1% | 25 | 2026-09-12 | independent | Source ↗ |
| CodeContests | Deprecated | 55.3% | 10 | 2026-09-12 | independent | Source ↗ |
| CodeXGLUE | Deprecated | 79.9% | 23 | 2026-09-12 | independent | Source ↗ |
| CommonsenseQA | Saturated | 97.7% | 28 | 2026-09-12 | independent | Source ↗ |
| ComplexBench | Nearing Saturation | 70.2% | 80 | 2026-09-12 | independent | Source ↗ |
| CyberSecEval | Deprecated | 76.5% | 10 | 2026-09-12 | independent | Source ↗ |
| FireBench | Active | 71.9% | 95 | 2026-09-12 | independent | Source ↗ |
| FrontierCode | Active | 49.3% | 98 | 2026-09-12 | independent | Source ↗ |
| FrontierMath | Active | 26% | 98 | 2026-09-12 | independent | Source ↗ |
| GLUE | Saturated | 100% | 28 | 2026-09-12 | independent | Source ↗ |
| GPQA | Nearing Saturation | 81.8% | 78 | 2026-09-12 | independent | Source ↗ |
| GSM-Hard | Deprecated | 89.3% | 10 | 2026-09-12 | independent | Source ↗ |
| GSM8K | Saturated | 100% | 25 | 2026-09-12 | independent | Source ↗ |
| HumanEval | Saturated | 96.9% | 25 | 2026-09-12 | independent | Source ↗ |
| HumanEval+ | Saturated | 91.8% | 25 | 2026-09-12 | independent | Source ↗ |
| Humanity's Last Exam | Active | 40.9% | 93 | 2026-09-12 | independent | Source ↗ |
| IFEval | Deprecated | 83.8% | 25 | 2026-09-12 | independent | Source ↗ |
| InfiniteBench | Nearing Saturation | 79.1% | 65 | 2026-09-12 | independent | Source ↗ |
| LiveCodeBench | Saturated | 69.7% | 39 | 2026-09-12 | independent | Source ↗ |
| LongBench | Nearing Saturation | 80% | 65 | 2026-09-12 | independent | Source ↗ |
| LongBench-Pro | Active | 76.3% | 80 | 2026-09-12 | independent | Source ↗ |
| MATH | Saturated | 91.1% | 25 | 2026-09-12 | independent | Source ↗ |
| MathArena | Active | 74.4% | 99 | 2026-09-12 | independent | Source ↗ |
| MBPP+ | Nearing Saturation | 86.7% | 65 | 2026-09-12 | independent | Source ↗ |
| LiveBench (Continuous) | Active | 80% | 94 | 2025-02-01 | independent | Source ↗ |
Local Hardware Execution (Ollama)Hardware Compatible
# 1. Pull and execute model locally with Ollama
ollama run qwen-3-8-max-prime
# 2. Or invoke via local OpenAI-compatible endpoint
curl http://localhost:11434/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "qwen-3-8-max-prime",
"messages": [{"role": "user", "content": "Analyze reasoning chains on AIME 2026"}]
}'Evaluation & Sourcing Notes
Scores listed for Qwen 3.8 Max Prime represent verified evaluations extracted from official research papers, independent evaluation suites (HELM, LMSYS, OpenCompass, LiveBench), and verified audit reports.
All benchmarks marked as saturated or deprecated reflect historical performance where the benchmark no longer provides active discriminative power.