Muse Spark 1.3 vs Muse Wonder Full
As of 18 Oct 2026, Muse Wonder Full (max) is the better pick for most workloads: it leads on quality (68.6 vs 67.2 on the BenchLeader Index).
Muse Spark 1.3 Key differences
- Muse Wonder Full (max) leads on quality: 68.6 vs 67.2.
- Muse Spark 1.3 (max) is stronger in agents & tools, coding, human preference, knowledge, maths, multimodal.
- Muse Wonder Full (max) is stronger in composite, long context, reasoning.
Choose Muse Spark 1.3 (max) if
- knowledge is the priority (73 vs 71 on the category score)
Choose Muse Wonder Full (max) if
- reasoning is the priority (95 vs 78 on the category score)
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Muse Spark 1.3 (max) | Muse Wonder Full (max) | Leader |
|---|---|---|---|
| Reasoning | 77.6 | 95.0 | Muse Wonder Full |
| Coding | 64.6 | 64.0 | level |
| Agents & tools | 65.6 | – | not comparable |
| Maths | 63.3 | – | not comparable |
| Knowledge | 72.7 | 71.3 | Muse Spark 1.3 |
| Human preference | 69.2 | – | not comparable |
| Multimodal | 65.9 | – | not comparable |
| Long context | 66.6 | 67.0 | level |
| Composite | 83.5 | 92.2 | Muse Wonder Full |
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | Muse Spark 1.3 (max) | Muse Wonder Full (max) |
|---|---|---|
| Blended price per 1M tokens | $2.00 | – |
| Input / output per 1M | $1.25 / $4.25 | – / – |
| Price per task: chat reply | $0.0018 | – |
| Price per task: summarise a 30-page report | $0.018 | – |
| Price per task: code edit | $0.014 | – |
| Price per task: agentic coding session | $0.092 | – |
| Price per task: structured extraction | $0.0034 | – |
| Cost to run a full benchmark suite | $1.60 | – |
| Output speed | 139 tok/s | – |
| Time to first answer | 67.7 s | – |
| Time to a complete answer | 71.3 s | – |
Every benchmark
| Metric | Muse Spark 1.3 (max) | Muse Wonder Full (max) |
|---|---|---|
| BenchLeader Index | 67.2 | 68.6 |
| Agents & tools score | 65.6 | – |
| Coding score | 64.6 | 64.0 |
| Composite score | 83.5 | 92.2 |
| Human preference score | 69.2 | – |
| Knowledge score | 72.7 | 71.3 |
| Long context score | 66.6 | 67.0 |
| Maths score | 63.3 | – |
| Multimodal score | 65.9 | – |
| Reasoning score | 77.6 | 95.0 |
| Blended price $/M | $2.00 | – |
| Output speed | 139 tok/s | – |
| Time to first answer | 67.7 s | – |
| Context window | 1.0M | 1M |
| FrontierMath Tiers 1–3 | 74.0% | – |
| FrontierMath Tier 4 | 46.3% | – |
| SciCode | 58.8% | – |
| APEX-Agents | 47.6% | – |
| ProofBench | 58.0% | – |
| Arena Text | 1494 | – |
| Arena Hard Prompts | 1515 | – |
| Arena Coding | 1535 | – |
| Arena WebDev | 1658 | – |
| Arena Vision | 1309 | – |
| Arena Agent | 3.8 | – |
| AA Intelligence Index v4.3.2 | 48.1 | 55.8 |
| AA-LCR | 83.0% | 83.7% |
| AA-Omniscience | 25 | 30.1 |
| GPQA Diamond (AA) | 93.5% | – |
| Humanity's Last Exam (AA) | 48.7% | 53.6% |
| SciCode (AA) | 58.8% | 62.6% |
| IOI | 56.6% | 79.8% |
| LegalBench | – | 87.4% |
| Terminal-Bench 2.1 (Vals) | 79.0% | – |
| Vals Index | 58.2 | 64.6 |
| CritPt | 24.9% | 32.9% |
| GDPval-AA v2.1 | 59.0% | 67.8% |
| τ³-Banking (AA) | 50.5% | – |
| ITBench SRE (AA) | 33.2% | – |
| Analyst Agent (AA) | 37.5% | – |
| BioMysteryBench | – | 73.3% |
| Code Migration | 47.4% | 61.0% |
| CUA-bench | 5.8% | – |
| CyberBench | 72.7% | 73.6% |
| Excel Modeling Benchmark | 67.4% | 71.9% |
| Finance Agent v2 | 60.0% | 59.1% |
| Harvey's Legal Agent Benchmark | 23.8% | 21.7% |
| Legal Research Bench | 55.3% | 53.4% |
| MedCode | – | 49.9% |
| MedScribe | – | 89.3% |
| MysteryMechanism | 36.0% | 35.1% |
| ProgramBench | 2.5% | 0.5% |
| Public Benefits Bench | – | 73.8% |
| SREBench | 3.8% | 11.8% |
| Tax Agent Bench | 42.4% | 43.0% |
| Terminal-Bench 4.0 (Vals) | 24.8% | 58.1% |
| Terminal-Bench Science | 10.0% | 37.1% |
| Vibe Code Bench 1-100 | 20.5% | 34.0% |
| Vibe Code Bench v1.1 | 85.9% | 86.0% |
| Arena Maths | 1504 | – |
| Arena Creative Writing | 1461 | – |
| Arena Instruction Following | 1483 | – |
| Arena Multi-turn | 1491 | – |
| Arena Longer Queries | 1496 | – |
| Arena Document | 1468 | – |
| Chess Puzzles | 37.0% | – |
| Mystery Game Puzzles | 25.0% | – |
| CursorBench | 41.6% | – |
| FrontierSWE | 25.8% | – |
| Terminal-Bench 4.0 (AA) | 33.3% | 57.1% |
| Terminal-Bench 2.1 (AA) | 84.3% | – |
| AutomationBench | 57.9% | 71.3% |
| GDP.pdf | 26.6% | 32.6% |
| MLCR | 43.3% | – |
| Harvey LAB-AA v1.1 | 8.9% | 13.1% |
| AA-Omniscience: accuracy | 43.6% | 43.2% |
| AA-Omniscience: non-hallucination | 67.1% | 77.0% |
| AA-Briefcase v1.1 | 1581 | 1729 |
| FrontierCode 1.1 (AA) | 40.6% | – |
| Terminal-Bench-Science 0.1 (AA) | 10.9% | – |
| AA-3D | 1037 | – |
Data as of 2026-10-18. Best configuration of each model; every score links to its source on the model pages.
Muse Spark 1.3 vs Muse Wonder Full: questions
- Is Muse Spark 1.3 better than Muse Wonder Full?
- Muse Wonder Full (max) leads on quality: 68.6 vs 67.2. The BenchLeader Index combines every independent quality benchmark; Muse Wonder Full (max) is ahead overall as of 2026-10-18, but check the category scores for your use.
- Is Muse Spark 1.3 better than Muse Wonder Full for coding?
- Muse Spark 1.3 scores higher in coding (65 vs 64 on the category index, where 50 is average).
- Which has the larger context window?
- Muse Spark 1.3 accepts more context: 1.0M against 1M tokens.