Muse Spark 1.2 vs Step 5 Preview
Full profiles: Muse Spark 1.2 (Meta) and Step 5 Preview (StepFun): every benchmark, price and rank.
Updated · Data from 9 sources
As of 16 Nov 2026, Muse Spark 1.2 is ahead of Step 5 Preview on quality (61.5 vs 56.0 on the BenchLeader Index), but Step 5 Preview is 1.4× cheaper at $1.43 vs $2.00 per 1M tokens blended. Choose Muse Spark 1.2 for reasoning (68 vs 52), knowledge (65 vs 54), and maths (67 vs 60) and fast first answers (11.4 s vs 23.5 s), where it also streams ~1.5× faster; choose Step 5 Preview for long context (69 vs 64).
Key facts (as of 16 Nov 2026)
- Muse Spark 1.2 scores 61.5 on the BenchLeader Index and Step 5 Preview scores 56.0 (BenchLeader, 16 Nov 2026).
- Muse Spark 1.2 costs $2.00 per 1M tokens blended (list price, 16 Nov 2026).
- Step 5 Preview costs $1.43 per 1M tokens blended (list price, 16 Nov 2026).
- Muse Spark 1.2 streams ~1.5× faster than Step 5 Preview (Artificial Analysis, 7-day median, 16 Nov 2026).
- Of the 35 benchmarks both models have, Step 5 Preview leads 18 (BenchLeader, 16 Nov 2026).
Which to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | Muse Spark 1.2 | 55 vs 50 category score; Arena Coding 1542 vs 1505 | Medium |
| Agents & tool use | Muse Spark 1.2 | 55 vs 49 category score; Arena Agent -8.2 vs -3.7 | Medium |
| Maths | Muse Spark 1.2 | 67 vs 60 category score; Arena Maths 1486 vs 1470 | Medium |
| Reasoning | Muse Spark 1.2 | 68 vs 52 category score; Arena Hard Prompts 1514 vs 1477 | High |
| Knowledge | Muse Spark 1.2 | 65 vs 54 category score; Public Benefits Bench v1.1 68.5% vs 58.1% | High |
| Long documents | Step 5 Preview | 69 vs 64 category score; AA-LCR v1.1 88.3% vs 79.0% | Medium |
| Images and other media | Muse Spark 1.2 | 65 vs 59 category score; Arena Vision 1305 vs 1263 | Medium |
| Chat quality (human preference) | Muse Spark 1.2 | 69 vs 63 on Arena votes | Medium |
| Lowest cost | Step 5 Preview | $1.43 vs $2.00 per 1M blended; agentic coding session $0.071 vs $0.092 | – |
| Fastest first answer | Muse Spark 1.2 | 11.4 s (Muse Spark 1.2 (xhigh)) vs 23.5 s (Step 5 Preview) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Muse Spark 1.2 (xhigh) leads on quality: 61.5 vs 56.0.
- Muse Spark 1.2 (xhigh) is stronger in agents & tools, coding, human preference, knowledge, maths, multimodal, reasoning.
- Step 5 Preview is stronger in long context.
- Step 5 Preview is 1.4× cheaper ($1.43 vs $2.00 per 1M blended).
- Muse Spark 1.2 (xhigh) streams 1.7× faster (208 vs 124 tokens per second).
At a glance
| Spec | Muse Spark 1.2 | Step 5 Preview |
|---|---|---|
| Developer | Meta | StepFun |
| Released | 5 Aug 2026 | 18 Sep 2026 |
| Context window (tokens) | 1,048,576 | 1,000,000 |
| Compared at | Muse Spark 1.2 (xhigh) | Step 5 Preview |
| Open weights | No | Yes |
| Price in / out / cached input per 1M | $1.25 / $4.25 / $0.150 | $1.00 / $2.70 / $0.105 |
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) | Step 5 Preview | Leader |
|---|---|---|---|
| Reasoning | 68.1 | 51.7 | Muse Spark 1.2 |
| Coding | 54.7 | 49.9 | Muse Spark 1.2 |
| Agents & tools | 55.4 | 49.1 | Muse Spark 1.2 |
| Maths | 66.8 | 60.1 | Muse Spark 1.2 |
| Knowledge | 64.5 | 53.7 | Muse Spark 1.2 |
| Human preference | 68.6 | 63.1 | Muse Spark 1.2 |
| Multimodal | 65.3 | 59.3 | Muse Spark 1.2 |
| Long context | 64.1 | 69.0 | Step 5 Preview |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Not drawn, measured for fewer than two of these models: Instruction following.
Where sources disagree
Across the 35 benchmarks both have, Muse Spark 1.2 leads 17, Step 5 Preview leads 18. Every overall index that rates both puts Muse Spark 1.2 ahead.
| Overall index | Muse Spark 1.2 (xhigh) | Step 5 Preview | Leader |
|---|---|---|---|
| BenchLeader Index | 61.5 | 56.0 | Muse Spark 1.2 |
| Arena Text | 1493 | 1449 | Muse Spark 1.2 |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
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. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | Muse Spark 1.2 (xhigh) | Step 5 Preview |
|---|---|---|
| Blended price per 1M tokens | $2.00 | $1.43 |
| Input / output per 1M | $1.25 / $4.25 | $1.00 / $2.70 |
| Cached input per 1M | $0.150 | $0.105 |
| Price per task: chat reply | $0.0018 | $0.0012 |
| Price per task: summarize a 30-page report | $0.018 | $0.014 |
| Price per task: code edit | $0.014 | $0.010 |
| Price per task: agentic coding session | $0.092 | $0.071 |
| Price per task: structured extraction | $0.0034 | $0.0025 |
| Agentic coding session, input mostly cached | $0.043 | $0.030 |
| Cost per benchmark task (AA Index) | – | $2.35 |
| Output tokens per task (AA Index) | – | 113k |
| Of which reasoning | – | 80k |
| Time per task (AA Index) | – | 15 min |
| Output speed | 208 tok/s | 124 tok/s |
| Time to first answer | 11.4 s | 23.5 s |
| Time to a complete answer, last measured 13 Nov 2026 | 23.9 s | 54.4 s |
Each model is compared at its best-ranked reasoning-effort setting. Muse Spark 1.2 (xhigh) scores 61.5 and takes 11 s to its first answer token. Step 5 Preview scores 56.0 and takes 24 s to its first answer token. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
126 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Muse Spark 1.2 (xhigh) | Step 5 Preview |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 61.5 | 56.0 |
| Agents & tools score | BenchLeader | 55.4 | 49.1 |
| Coding score | BenchLeader | 54.7 | 49.9 |
| Human preference score | BenchLeader | 68.6 | 63.1 |
| Knowledge score | BenchLeader | 64.5 | 53.7 |
| Long context score | BenchLeader | 64.1 | 69.0 |
| Maths score | BenchLeader | 66.8 | 60.1 |
| Multimodal score | BenchLeader | 65.3 | 59.3 |
| Reasoning score | BenchLeader | 68.1 | 51.7 |
| Blended price $/M | BenchLeader | $2.00 | $1.43 |
| Output speed | BenchLeader | 208 tok/s | 124 tok/s |
| Time to first answer | BenchLeader | 11.4 s | 23.5 s |
| Context window | BenchLeader | 1,048,576 tokens | 1,000,000 tokens |
| Reasoning | |||
| SimpleBench | SimpleBench | 74.5%† | – |
| Arena Hard Prompts | Arena (LMArena) | 1514 | 1477 |
| LiveBench 2.0 Reasoning | LiveBench | – | 83.8%† |
| LiveBench 1.0 Reasoning | LiveBench | 90.0% | – |
| GPQA Diamond (AA) | Artificial Analysis | 90.4% | – |
| Humanity's Last Exam (AA) | Artificial Analysis | 45.5% | 46.5% |
| HLE-Diamond (AA) | Artificial Analysis | – | 24.3% |
| CritPt | Artificial Analysis | 17.7% | 20.9% |
| CritPt v2 hard subset (AA) | Artificial Analysis | – | 32.5% |
| MysteryMechanism | Vals AI | – | 18.5% |
| Arena Expert | Arena (LMArena) | 1503 | 1489 |
| Chess Puzzles | Epoch AI Benchmarking Hub | 27.0% | – |
| LMCA | Epoch AI Benchmarking Hub | 48.4% | – |
| DTBench | Epoch AI Benchmarking Hub | 94.7% | – |
| Coding | |||
| SciCode | SciCode | 57.4% | 58.9% |
| WeirdML | WeirdML | 60.3% | – |
| Arena Coding | Arena (LMArena) | 1542 | 1505 |
| Arena WebDev | Arena (LMArena) | 1534 | 1573† |
| LiveBench 2.0 Coding | LiveBench | – | 67.6%† |
| LiveBench 1.0 Coding | LiveBench | 77.5% | – |
| IOI | Vals AI | 21.8% | 38.1% |
| SWE-bench (Vals) | Vals AI | 86.6% | – |
| Code Migration | Vals AI | 29.9% | 39.0% |
| Code Migration v1.1 | Vals AI | – | 47.0% |
| ProgramBench | Vals AI | 0.5% | – |
Show all 126 benchmarks
| Metric | Publisher | Muse Spark 1.2 (xhigh) | Step 5 Preview |
|---|---|---|---|
| Coding | |||
| SWE-bench Multimodal | Vals AI | – | 40.8% |
| Vibe Code Bench v1.1 | Vals AI | 79.1% | 70.4% |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 54.9% | – |
| FrontierSWE | Epoch AI Benchmarking Hub | 12.0% | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 7.1% | 33.3% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 80.2% | – |
| Startrise Frontend | Startrise | 70.3%† | – |
| GameBench 2 Overall | DuelLab GameBench 2 | 51.5† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 56.9 | – |
| Design Arena Website | Design Arena | 1309† | – |
| FrontierCode 1.1 (AA) | Artificial Analysis | – | 38.4% |
| FrontierCode 2.0 (AA) | Artificial Analysis | – | 12.0% |
| Agents & tools | |||
| APEX-Agents | Mercor | 45.4%† | – |
| Arena Agent | Arena (LMArena) | -8.2 | -3.7 |
| LiveBench Long Agentic | LiveBench | – | 19.4%† |
| LiveBench Browser | LiveBench | – | 24.8%† |
| LiveBench Agentic Coding | LiveBench | 57.6% | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | 69.7% | – |
| GDPval-AA v2.1, normalised score | Artificial Analysis | – | 54.3% |
| GDPval-AA v2.1 | Artificial Analysis | 1478 | 1591 |
| τ³-Banking (AA) | Artificial Analysis | 34.9% | – |
| ITBench SRE (AA) | Artificial Analysis | – | 55.6% |
| Analyst Agent (AA) | Artificial Analysis | – | 35.0% |
| APEX-Agents (AA) | Artificial Analysis | – | 38.0% |
| CyberBench | Vals AI | 69.5% | – |
| Finance Agent v2 | Vals AI | 60.6% | 50.7% |
| Finance Agent v2.1 | Vals AI | – | 58.9% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | – | 12.9% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 25.4% | 9.2% |
| Legal Research Bench | Vals AI | 43.8% | 40.4% |
| SkillsBench | Vals AI | 53.0% | – |
| SREBench | Vals AI | 0.0% | 0.4% |
| Tax Agent Bench | Vals AI | 32.8% | 27.3% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 6.1% | 31.8% |
| Terminal-Bench Science | Vals AI | – | 2.9% |
| Arena Agent Task Completion | Arena (LMArena) | -9.8 | -1.5 |
| AutomationBench | Artificial Analysis | 40.6% | 51.0% |
| GDP.pdf (AA) | Artificial Analysis | 17.4% | 14.8% |
| EnterpriseOps-Gym | Artificial Analysis | 47.3% | 47.2% |
| AA-Briefcase v1.1 | Artificial Analysis | 1323 | 1424 |
| Design Arena Agentic | Design Arena | 1185† | – |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 2.4% |
| CWE-Bench (AA) | Artificial Analysis | – | 52.5% |
| DeepSec-Bench (AA) | Artificial Analysis | – | 15.8% |
| CyberGym end-to-end (AA) | Artificial Analysis | – | 74.8% |
| MKA (AA) | Artificial Analysis | – | 27.6% |
| Maths | |||
| ProofBench | Vals AI | 43.0%† | 42.0% |
| LiveBench Mathematics | LiveBench | 91.2% | – |
| Arena Maths | Arena (LMArena) | 1486 | 1470 |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 60.3% | – |
| LiveBench 2.0 Data Analysis | LiveBench | – | 63.5%† |
| LiveBench Research | LiveBench | – | 75.5%† |
| LiveBench 1.0 Data Analysis | LiveBench | 76.5% | – |
| AA-Omniscience v1.1 | Artificial Analysis | – | 16.4 |
| MMLU-Pro | Vals AI | 88.3% | – |
| LegalBench | Vals AI | 85.3% | – |
| CorpFin | Vals AI | 70.9% | – |
| TaxEval | Vals AI | 80.4% | – |
| BioMysteryBench | Vals AI | 64.8% | 67.4% |
| Excel Modeling Benchmark | Vals AI | 57.0% | 57.1% |
| Excel Modeling Benchmark v1.1 | Vals AI | – | 69.5% |
| MedCode | Vals AI | 49.4% | – |
| MedScribe | Vals AI | 90.1% | – |
| Public Benefits Bench v1.1 | Vals AI | 68.5% | 58.1% |
| Financial Coaching Bench | Vals AI | 75.7% | – |
| Immigration Bench | Vals AI | 66.5% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 16.0% | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | – | 36.4% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | – | 68.7% |
| Instruction following | |||
| LiveBench Language | LiveBench | 78.6% | – |
| LiveBench Instruction Following | LiveBench | 74.3% | – |
| Arena Instruction Following | Arena (LMArena) | 1479 | 1450 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1493 | 1449 |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1842† | – |
| Arena Creative Writing | Arena (LMArena) | 1443 | 1413 |
| Arena Multi-turn | Arena (LMArena) | 1510 | 1445 |
| Arena Longer Queries | Arena (LMArena) | 1491 | 1461 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1305 | 1263 |
| LiveBench Design | LiveBench | – | 70.8%† |
| MMMU-Pro | Artificial Analysis | – | 76.4% |
| MMMU-Pro (Vals) | Vals AI | 86.1% | – |
| MortgageTax | Vals AI | 65.4% | – |
| Humanity's Sixth Sense | Scale AI SEAL | 24.5% | – |
| Code2Video-AA | Artificial Analysis | – | 821 |
| AA-3D | Artificial Analysis | – | 859 |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 79.0% | 88.3% |
| MLCR | Artificial Analysis | 7.2% | 18.3% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 155.0† | – |
| LiveBench 2.0 | LiveBench | – | 54.9%† |
| LiveBench 1.0 | LiveBench | 78.0% | – |
| AA Intelligence Index v5.0.1 | Artificial Analysis | – | 28.5 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 39.6 | 43.7 |
| AA Progress Index v1 | Artificial Analysis | 309.1 | 318.2 |
| Vals Index | Vals AI | – | 54.8 |
| Vals Finance Index | Vals AI | – | 55.4% |
| Vals Science Index | Vals AI | – | 29.6% |
| Vals Multimodal Index | Vals AI | 69.8%† | – |
| Vals Legal Index | Vals AI | – | 70.2% |
| AA Cyber Index | Artificial Analysis | – | 47.7 |
| Safety & honesty | |||
| DistressBench | Scale AI SEAL | 88.4%† | – |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 33.3 |
Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- Step 5 Preview: BenchLeader Index 54.2 → 56.0.
- Step 5 Preview: BenchLeader Index 54.7 → 54.2.
- Muse Spark 1.2 (xhigh): 1 new result (GDPval-AA v2.1).
- Step 5 Preview: 1 score revision.
- Step 5 Preview: 7 score revisions.
- Muse Spark 1.2 (xhigh): 1 score revision.
- Step 5 Preview: 2 new results (Vals Index, Harvey's Legal Agent Benchmark v1.1).
- Muse Spark 1.2 (xhigh): 1 score revision.
- Step 5 Preview: 1 score revision.
- Muse Spark 1.2 (xhigh): BenchLeader Index 62.2 → 61.6.
Related comparisons
- Muse Spark 1.2 vs Claude Fable 5.2
- Muse Spark 1.2 vs Gemini 4 Carbon
- Muse Spark 1.2 vs Claude Opus 5.5
- Step 5 Preview vs Claude Opus 5.5
- Step 5 Preview vs GPT-6 Astra
- Step 5 Preview vs GPT-6.1 Sol
- Step 5 Preview vs Gemini 4 Argon
- Muse Spark 1.1 vs Step 5 Preview
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Muse Spark 1.2 vs Step 5 Preview: questions
- Is Muse Spark 1.2 better than Step 5 Preview?
- Muse Spark 1.2 (xhigh) leads on quality: 61.5 vs 56.0. The BenchLeader Index combines every independent quality benchmark; Muse Spark 1.2 (xhigh) is ahead overall as of 16 Nov 2026, but check the category scores for your use.
- Muse Spark 1.2 vs Step 5 Preview: which is cheaper, and will that change?
- Step 5 Preview is cheaper today: $1.43 against $2.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Muse Spark 1.2 or Step 5 Preview better for coding?
- Muse Spark 1.2 scores higher in coding (55 vs 50 on the category index, where 50 is average). On SciCode, the top shared coding benchmark, it is 57.4% for Muse Spark 1.2 against 58.9% for Step 5 Preview.
- Which is faster, Muse Spark 1.2 or Step 5 Preview?
- Muse Spark 1.2 streams ~1.5× faster (208 vs 124 output tokens per second, Artificial Analysis 7-day median).
- Is Muse Spark 1.2 or Step 5 Preview better for agents and tool use?
- Muse Spark 1.2 scores higher in agentic tasks (55 vs 49 on the category index, where 50 is average). On Arena Agent, the top shared agents & tools benchmark, it is -8.2 for Muse Spark 1.2 against -3.7 for Step 5 Preview.
More questions (4)
- Which is better for long documents?
- Step 5 Preview scores higher on long-context benchmarks (69 vs 64). Both take about 1M tokens: Muse Spark 1.2 1,048,576, Step 5 Preview 1,000,000.
- Which has lower latency?
- Muse Spark 1.2 answers first: 11.4 s against 23.5 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
- Do independent benchmarks agree on Muse Spark 1.2 vs Step 5 Preview?
- Across the 35 benchmarks both have, Muse Spark 1.2 leads 17, Step 5 Preview leads 18. Every overall index that rates both puts Muse Spark 1.2 ahead.
- When was Step 5 Preview released?
- Step 5 Preview was released on 18 Sep 2026.