Muse Spark 1.2 vs Muse Wonder
As of 13 Oct 2026, Muse Wonder (max) leads on quality (66.3 vs 62.6 on the BenchLeader Index), but Muse Spark 1.2 (xhigh) costs 4.0× less, so it is the better buy unless you need the extra quality.
Muse Spark 1.2 Key differences
- Muse Wonder (max) leads on quality: 66.3 vs 62.6.
- Muse Spark 1.2 (xhigh) is stronger in agents & tools, coding, human preference, maths, multimodal.
- Muse Wonder (max) is stronger in composite, knowledge, long context, reasoning.
- Muse Spark 1.2 (xhigh) is 4.0× cheaper ($2.00 vs $8.00 per 1M blended).
- Muse Spark 1.2 (xhigh) streams 3.2× faster (272 vs 85 tokens per second).
Choose Muse Spark 1.2 (xhigh) if
- cost matters: $2.00 against $8.00 per million tokens blended
- you need fast output: 272 against 85 tokens per second
- the wait for the first word matters: 20.4 s against 60.8 s to first answer
Choose Muse Wonder (max) if
- reasoning is the priority (88 vs 70 on the category score)
- knowledge is the priority (73 vs 67 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 1.2 (xhigh) | Muse Wonder (max) | Leader |
|---|---|---|---|
| Reasoning | 69.5 | 87.8 | Muse Wonder |
| Coding | 55.0 | – | not comparable |
| Agents & tools | 55.9 | – | not comparable |
| Maths | 66.7 | – | not comparable |
| Knowledge | 66.9 | 73.0 | Muse Wonder |
| Human preference | 69.1 | – | not comparable |
| Multimodal | 65.3 | – | not comparable |
| Long context | 64.6 | 65.5 | level |
| Composite | 66.3 | 78.2 | Muse Wonder |
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 1.2 (xhigh) | Muse Wonder (max) |
|---|---|---|
| Blended price per 1M tokens | $2.00 | $8.00 |
| Input / output per 1M | $1.25 / $4.25 | $4.00 / $20.00 |
| Price per task: chat reply | $0.0018 | $0.0076 |
| Price per task: summarise a 30-page report | $0.018 | $0.060 |
| Price per task: code edit | $0.014 | $0.054 |
| Price per task: agentic coding session | $0.092 | $0.320 |
| Price per task: structured extraction | $0.0034 | $0.012 |
| Cost to run a full benchmark suite | $0.975 | $4.52 |
| Output speed | 272 tok/s | 85 tok/s |
| Time to first answer | 20.4 s | 60.8 s |
| Time to a complete answer | 22.3 s | 66.7 s |
Every benchmark
| Metric | Muse Spark 1.2 (xhigh) | Muse Wonder (max) |
|---|---|---|
| BenchLeader Index | 62.6 | 66.3 |
| Agents & tools score | 55.9 | – |
| Coding score | 55.0 | – |
| Composite score | 66.3 | 78.2 |
| Human preference score | 69.1 | – |
| Knowledge score | 66.9 | 73.0 |
| Long context score | 64.6 | 65.5 |
| Maths score | 66.7 | – |
| Multimodal score | 65.3 | – |
| Reasoning score | 69.5 | 87.8 |
| Blended price $/M | $2.00 | $8.00 |
| Output speed | 272 tok/s | 85 tok/s |
| Time to first answer | 20.4 s | 60.8 s |
| Context window | 1.0M | 1.0M |
| SimpleQA Verified | 60.3% | – |
| SciCode | 56.4% | – |
| WeirdML | 60.3% | – |
| LMArena Text | 1492 | – |
| LMArena Hard Prompts | 1508 | – |
| LMArena Coding | 1531 | – |
| LMArena WebDev | 1533 | – |
| LMArena Vision | 1305 | – |
| LMArena Agent | -3.7 | – |
| LiveBench | 78.0% | 81.2% |
| LiveBench Reasoning | 90.0% | 90.6% |
| LiveBench Coding | 77.5% | 80.7% |
| LiveBench Agentic Coding | 57.6% | 62.7% |
| LiveBench Mathematics | 91.2% | 96.5% |
| LiveBench Data Analysis | 76.5% | 81.6% |
| LiveBench Language | 78.6% | 82.9% |
| LiveBench Instruction Following | 74.3% | 73.0% |
| AA Intelligence Index v4.3.2 | 39.6 | 52.4 |
| AA-LCR | 79.0% | 80.7% |
| AA-Omniscience | 27.2 | 25.5 |
| GPQA Diamond (AA) | 90.4% | – |
| Humanity's Last Exam (AA) | 45.5% | 50.1% |
| SciCode (AA) | 57.4% | 60.3% |
| MMLU-Pro | 88.3% | – |
| IOI | 21.8% | – |
| LegalBench | 85.3% | – |
| CorpFin | 70.9% | – |
| TaxEval | 80.4% | – |
| Terminal-Bench 2.1 (Vals) | 69.7% | – |
| SWE-bench (Vals) | 86.6% | – |
| Vals Index | 49.3 | – |
| CritPt | 17.7% | 28.3% |
| GDPval-AA v2.1 | 48.7% | 63.5% |
| τ³-Banking (AA) | 34.9% | – |
| BioMysteryBench | 64.8% | – |
| Code Migration | 29.9% | – |
| CyberBench | 69.5% | – |
| Excel Modeling Benchmark | 57.0% | – |
| Finance Agent v2 | 60.6% | – |
| Harvey's Legal Agent Benchmark | 25.4% | – |
| Legal Research Bench | 43.8% | – |
| MedCode | 49.4% | – |
| MedScribe | 90.1% | – |
| MMMU-Pro (Vals) | 86.1% | – |
| MortgageTax | 65.4% | – |
| Public Benefits Bench | 68.5% | – |
| SAGE | 47.7% | – |
| SkillsBench | 53.0% | – |
| Tax Agent Bench | 56.9% | – |
| Terminal-Bench 4.0 (Vals) | 6.1% | – |
| Vibe Code Bench v1.1 | 79.1% | – |
| LMArena Maths | 1481 | – |
| LMArena Creative Writing | 1454 | – |
| LMArena Instruction Following | 1471 | – |
| LMArena Multi-turn | 1506 | – |
| LMArena Longer Queries | 1489 | – |
| DeepSWE v1.1 | 54.9% | – |
| LMCA | 48.4% | – |
| DTBench | 94.7% | – |
| GDP.pdf | 16.0% | – |
| FrontierSWE | 12.0% | – |
| Terminal-Bench 4.0 (AA) | 7.1% | 52.0% |
| Terminal-Bench 2.1 (AA) | 80.2% | – |
| AutomationBench | – | 68.1% |
| GDP.pdf | – | 30.6% |
| AA-Omniscience: accuracy | 45.4% | 41.1% |
| AA-Omniscience: non-hallucination | 66.7% | 73.5% |
| AA-Briefcase v1.1 | – | 1659 |
Data as of 2026-10-13. Best configuration of each model; every score links to its source on the model pages.
Muse Spark 1.2 vs Muse Wonder: questions
- Is Muse Spark 1.2 better than Muse Wonder?
- Muse Wonder (max) leads on quality: 66.3 vs 62.6. The BenchLeader Index combines every independent quality benchmark; Muse Wonder (max) is ahead overall as of 2026-10-13, but check the category scores for your use.
- Which is cheaper, Muse Spark 1.2 or Muse Wonder?
- Muse Spark 1.2 is cheaper: $2.00 against $8.00 per million tokens, blended at three input tokens per output token.
- Which is faster, Muse Spark 1.2 or Muse Wonder?
- Muse Spark 1.2 streams faster: 272 against 85 output tokens per second.
- Which has the larger context window?
- Both accept 1.0M tokens of context.