Meta AIproprietary
Released: 2025-04-15Muse Wonder Full
Meta flagship high-capacity autonomous agent model powering premium Muse Code tiers and standalone Muse agent app, delivering top-tier performance on DeepSWE and SWE-bench Verified.
Intelligence Index
95.2/ 100
Calibrated multi-domain compositeThroughput Speed
65tok/sec
Output streaming throughputTime To First Token
440ms
Initial chunk server latencyLiveBench Contam-Free
83.5%
September 2026 suite scoreSourced Benchmark Evaluations (61)
Verified performance across authoritative benchmarks with provenance tracking.| Benchmark | Status | Score | Trust | Date | Source Type | Provenance |
|---|---|---|---|---|---|---|
| ARC-AGI-3 | Active | 61.8% | 93 | 2025-05-20 | independent | Source ↗ |
| ARC-AGI-2 | Nearing Saturation | 74.1% | 75 | 2025-05-13 | independent | Source ↗ |
| FrontierMath | Active | 26.5% | 98 | 2025-05-10 | independent | Source ↗ |
| PutnamBench | Active | 49.1% | 90 | 2025-05-09 | independent | Source ↗ |
| SWE-Lancer | Deprecated | 61.9% | 20 | 2025-05-07 | independent | Source ↗ |
| Omni-MATH | Deprecated | 69.9% | 10 | 2025-05-07 | independent | Source ↗ |
| Soohak | Active | 71.8% | 98 | 2025-05-05 | independent | Source ↗ |
| FrontierCode | Active | 52.8% | 98 | 2025-05-05 | independent | Source ↗ |
| Humanity's Last Exam | Active | 41.8% | 93 | 2025-05-03 | independent | Source ↗ |
| StructFlowBench | Nearing Saturation | 74.4% | 75 | 2025-05-03 | independent | Source ↗ |
| AIME | Nearing Saturation | 89.8% | 75 | 2025-05-03 | independent | Source ↗ |
| OlympiadBench | Active | 77.5% | 80 | 2025-05-02 | independent | Source ↗ |
| MathArena | Active | 75.6% | 99 | 2025-05-01 | independent | Source ↗ |
| FireBench | Active | 73.6% | 95 | 2025-05-01 | independent | Source ↗ |
| Terminal-Bench | Active | 63.7% | 80 | 2025-05-01 | independent | Source ↗ |
| OpenAI-MRCR | Active | 80% | 80 | 2025-04-30 | independent | Source ↗ |
| SimpleQA | Active | 49.5% | 90 | 2025-04-30 | independent | Source ↗ |
| CodeContests | Deprecated | 59.2% | 10 | 2025-04-30 | independent | Source ↗ |
| ComplexBench | Nearing Saturation | 71.8% | 80 | 2025-04-29 | independent | Source ↗ |
| SWE-bench Pro | Active | 56.4% | 90 | 2025-04-29 | independent | Source ↗ |
| LongBench-Pro | Active | 78.1% | 80 | 2025-04-29 | independent | Source ↗ |
| ZeroSCROLLS | Nearing Saturation | 76.2% | 78 | 2025-04-28 | independent | Source ↗ |
| MultiPL-E | Deprecated | 83.7% | 10 | 2025-04-28 | independent | Source ↗ |
| LiveCodeBench | Saturated | 74.6% | 39 | 2025-04-27 | independent | Source ↗ |
| AgentIF | Nearing Saturation | 59.6% | 80 | 2025-04-27 | independent | Source ↗ |
| MATH | Saturated | 92.6% | 25 | 2025-04-27 | independent | Source ↗ |
| InfiniteBench | Nearing Saturation | 80.9% | 65 | 2025-04-27 | independent | Source ↗ |
| Mercury | Deprecated | 80.1% | 10 | 2025-04-26 | independent | Source ↗ |
| SCROLLS | Nearing Saturation | 76.2% | 65 | 2025-04-26 | independent | Source ↗ |
| MMLU-Pro | Nearing Saturation | 87.6% | 75 | 2025-04-26 | independent | Source ↗ |
| RULER | Nearing Saturation | 87.6% | 84 | 2025-04-25 | independent | Source ↗ |
| TabMWP | Saturated | 86.9% | 25 | 2025-04-25 | independent | Source ↗ |
| CyberSecEval | Deprecated | 81.9% | 10 | 2025-04-25 | independent | Source ↗ |
| SWE-bench Verified | Saturated | 71% | 25 | 2025-04-24 | independent | Source ↗ |
| MMLU-Redux | Saturated | 93.3% | 25 | 2025-04-24 | independent | Source ↗ |
| QuALITY | Nearing Saturation | 83.8% | 78 | 2025-04-24 | independent | Source ↗ |
| LongBench | Nearing Saturation | 81.9% | 65 | 2025-04-23 | independent | Source ↗ |
| SWE-bench | Deprecated | 61.9% | 10 | 2025-04-23 | independent | Source ↗ |
| MGSM | Nearing Saturation | 89.8% | 65 | 2025-04-23 | independent | Source ↗ |
| AGIEval | Saturated | 81.7% | 25 | 2025-04-22 | independent | Source ↗ |
| QASPER | Nearing Saturation | 79% | 65 | 2025-04-22 | independent | Source ↗ |
| TruthfulQA | Saturated | 83.8% | 25 | 2025-04-22 | independent | Source ↗ |
| GSM-Hard | Deprecated | 90.7% | 10 | 2025-04-21 | independent | Source ↗ |
| SuperGLUE | Saturated | 96.9% | 25 | 2025-04-21 | independent | Source ↗ |
| CodeXGLUE | Deprecated | 85.5% | 23 | 2025-04-20 | independent | Source ↗ |
| MT-Bench | Nearing Saturation | 9% | 75 | 2025-04-20 | independent | Source ↗ |
| BIG-Bench Hard | Saturated | 93.1% | 25 | 2025-04-20 | independent | Source ↗ |
| WinoGrande | Saturated | 98.8% | 28 | 2025-04-19 | independent | Source ↗ |
| MBPP+ | Nearing Saturation | 92.8% | 65 | 2025-04-19 | independent | Source ↗ |
| IFEval | Deprecated | 85.8% | 25 | 2025-04-19 | independent | Source ↗ |
| GPQA | Nearing Saturation | 83.8% | 78 | 2025-04-19 | independent | Source ↗ |
| Needle-in-a-Haystack | Saturated | 99% | 44 | 2025-04-18 | independent | Source ↗ |
| HumanEval+ | Saturated | 98.3% | 25 | 2025-04-18 | independent | Source ↗ |
| CommonsenseQA | Saturated | 99.8% | 28 | 2025-04-18 | independent | Source ↗ |
| MBPP | Saturated | 96.5% | 25 | 2025-04-17 | independent | Source ↗ |
| GSM8K | Saturated | 100% | 25 | 2025-04-17 | independent | Source ↗ |
| GLUE | Saturated | 100% | 28 | 2025-04-17 | independent | Source ↗ |
| MMLU | Saturated | 97.1% | 25 | 2025-04-16 | independent | Source ↗ |
| HumanEval | Saturated | 100% | 25 | 2025-04-16 | independent | Source ↗ |
| ARC-AGI | Saturated | 89.3% | 25 | 2025-04-16 | independent | Source ↗ |
| LiveBench (Continuous) | Active | 83.5% | 94 | 2025-02-01 | independent | Source ↗ |
OpenAI Compatible API ExecutionAPI Ready
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="muse-wonder-full",
messages=[{"role": "user", "content": "Evaluate multi-step logic problem"}],
temperature=0.2
)
print(response.choices[0].message.content)Evaluation & Sourcing Notes
Scores listed for Muse Wonder Full 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.