GPT-5.4 vs Muse Spark 1.2
Full profiles: GPT-5.4 (OpenAI) and Muse Spark 1.2 (Meta): every benchmark, price and rank.
Updated · Data from 11 sources
As of 25 Oct 2026, GPT-5.4 (xhigh) and Muse Spark 1.2 (xhigh) are neck and neck on quality (64.7 vs 63.2 on the BenchLeader Index, inside the ±5.3 margin of error), and Muse Spark 1.2 (xhigh) costs 2.8× less, which makes it the better buy for most workloads. Pick GPT-5.4 for coding (65 vs 55); pick Muse Spark 1.2 for knowledge (67 vs 60).
GPT-5.4 benchmarks 
Muse Spark 1.2 benchmarks 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 | GPT-5.4 | 65 vs 55 category score; WeirdML 77.7% vs 60.3% | High |
| Agents & tool use | GPT-5.4 | 64 vs 55 category score; τ³-Banking (AA) 39.6% vs 34.9% | High |
| Maths | Muse Spark 1.2 | 67 vs 65 category score; LiveBench Mathematics 91.2% vs 94.2% | Medium |
| Reasoning | Either | 69 vs 69 category score; CritPt 23.4% vs 17.7% | Either |
| Knowledge | Muse Spark 1.2 | 67 vs 60 category score; SimpleQA Verified 60.3% vs 45.1% | Medium |
| Long documents | GPT-5.4 | 66 vs 65 category score; AA-LCR 82.0% vs 79.0% | Medium |
| Images and other media | Muse Spark 1.2 | 65 vs 61 category score; SAGE 47.7% vs 43.3% | Medium |
| Chat quality (human preference) | – | Only Muse Spark 1.2 has an Arena rating (69 category score) | Not comparable |
| Lowest cost | Muse Spark 1.2 | $2.00 vs $5.63 per 1M blended; agentic coding session $0.092 vs $0.210 | – |
| Fastest first answer | Muse Spark 1.2 | 17.6 s (Muse Spark 1.2 (xhigh)) vs 118.0 s (GPT-5.4 (xhigh)) 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
- GPT-5.4 (xhigh) and Muse Spark 1.2 (xhigh) are neck and neck on quality (64.7 vs 63.2, within the ±5.3 margin).
- GPT-5.4 (xhigh) is stronger in agents & tools, coding, instruction following, long context.
- Muse Spark 1.2 (xhigh) is stronger in composite, knowledge, maths, multimodal, reasoning, human preference.
- Muse Spark 1.2 (xhigh) is 2.8× cheaper ($2.00 vs $5.63 per 1M blended).
- Muse Spark 1.2 (xhigh) streams 3.2× faster (307 vs 96 tokens per second).
At a glance
| Spec | GPT-5.4 | Muse Spark 1.2 |
|---|---|---|
| Developer | OpenAI | Meta |
| Released | 5 Mar 2026 | 5 Aug 2026 |
| Context window (tokens) | 1,050,000 | 1,048,576 |
| Compared at | GPT-5.4 (xhigh) | Muse Spark 1.2 (xhigh) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $2.50 / $15.00 / $0.250 | $1.25 / $4.25 / $0.150 |
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 | GPT-5.4 (xhigh) | Muse Spark 1.2 (xhigh) | Leader |
|---|---|---|---|
| Reasoning | 69.1 | 69.3 | level |
| Coding | 64.6 | 55.1 | GPT-5.4 |
| Agents & tools | 63.6 | 55.4 | GPT-5.4 |
| Maths | 65.2 | 67.3 | Muse Spark 1.2 |
| Knowledge | 59.7 | 66.7 | Muse Spark 1.2 |
| Instruction following | 72.3 | – | not comparable |
| Human preference | – | 68.6 | not comparable |
| Multimodal | 61.4 | 65.2 | Muse Spark 1.2 |
| Long context | 66.1 | 64.5 | GPT-5.4 |
| Composite | 73.1 | 73.8 | level |
Where sources disagree
Across the 39 benchmarks both have, GPT-5.4 leads 16, Muse Spark 1.2 leads 22 and 1 is a tie. Every overall index that rates both puts Muse Spark 1.2 ahead.
| Overall index | GPT-5.4 (xhigh) | Muse Spark 1.2 (xhigh) | Leader |
|---|---|---|---|
| BenchLeader Index | 64.7 | 63.2 | neck and neck |
| AA Intelligence Index v4.3.2 | 39.0 | 39.6 | Muse Spark 1.2 |
| Arena Text | 1475† | 1491 | 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; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | GPT-5.4 (xhigh) | Muse Spark 1.2 (xhigh) |
|---|---|---|
| Blended price per 1M tokens | $5.63 | $2.00 |
| Input / output per 1M | $2.50 / $15.00 | $1.25 / $4.25 |
| Cached input per 1M | $0.250 | $0.150 |
| Price per task: chat reply | $0.0055 | $0.0018 |
| Price per task: summarize a 30-page report | $0.039 | $0.018 |
| Price per task: code edit | $0.037 | $0.014 |
| Price per task: agentic coding session | $0.210 | $0.092 |
| Price per task: structured extraction | $0.0080 | $0.0034 |
| Cost to run a full benchmark suite | – | $0.975 |
| Output speed | 96 tok/s | 307 tok/s |
| Time to first answer | 118.0 s | 17.6 s |
| Time to a complete answer | 123.2 s | 19.2 s |
Each model is compared at its best-ranked reasoning-effort setting. GPT-5.4 (xhigh) scores 64.7 and takes 118 s to its first answer token; GPT-5.4 (low) scores 59.7 in 2 s, GPT-5.4 (high) scores 58.9 in 1 s. Muse Spark 1.2 (xhigh) scores 63.2 and takes 18 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
144 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 | GPT-5.4 (xhigh) | Muse Spark 1.2 (xhigh) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 64.7 | 63.2 |
| Agents & tools score | BenchLeader | 63.6 | 55.4 |
| Coding score | BenchLeader | 64.6 | 55.1 |
| Composite score | BenchLeader | 73.1 | 73.8 |
| Instruction following score | BenchLeader | 72.3 | – |
| Knowledge score | BenchLeader | 59.7 | 66.7 |
| Long context score | BenchLeader | 66.1 | 64.5 |
| Maths score | BenchLeader | 65.2 | 67.3 |
| Multimodal score | BenchLeader | 61.4 | 65.2 |
| Reasoning score | BenchLeader | 69.1 | 69.3 |
| Human preference score | BenchLeader | – | 68.6 |
| Blended price $/M | BenchLeader | $5.63 | $2.00 |
| Output speed | BenchLeader | 96 tok/s | 307 tok/s |
| Time to first answer | BenchLeader | 118.0 s | 17.6 s |
| Context window | BenchLeader | 1,050,000 tokens | 1,048,576 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 93.3% | – |
| Humanity's Last Exam | Scale AI / CAIS | 36.2% | – |
| SimpleBench | SimpleBench | – | 74.5%† |
| Arena Hard Prompts | Arena (LMArena) | 1487† | 1512 |
| LiveBench 2.0 Reasoning | LiveBench | – | 90.0%† |
| LiveBench 1.0 Reasoning | LiveBench | 88.1% | 90.0% |
| GPQA Diamond (AA) | Artificial Analysis | 92.0% | 90.4% |
| Humanity's Last Exam (AA) | Artificial Analysis | 43.7% | 45.5% |
| GPQA Diamond (Vals) | Vals AI | 91.7% | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 63.8%† | – |
| ARC-AGI-1 | ARC Prize | 93.7% | – |
| ARC-AGI-2 | ARC Prize | 74.0% | – |
| ARC-AGI-3 | ARC Prize | 0.2%† | – |
| CritPt | Artificial Analysis | 23.4% | 17.7% |
| Chess Puzzles | Epoch AI Benchmarking Hub | 48.0% | 27.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 37.0% | – |
| CL-bench | Epoch AI Benchmarking Hub | 27.9% | – |
| CL-bench Life | Epoch AI Benchmarking Hub | 21.7% | – |
| LMCA | Epoch AI Benchmarking Hub | 52.0% | 48.4% |
| DTBench | Epoch AI Benchmarking Hub | 94.4% | 94.7% |
| ForecastBench | ForecastBench | 63.1%† | – |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 76.9%† | – |
| SciCode | SciCode | 56.6% | 57.4% |
| WeirdML | WeirdML | 77.7% | 60.3% |
| GSO-Bench | GSO-Bench | 31.4% | – |
Show all 144 benchmarks
| Metric | Publisher | GPT-5.4 (xhigh) | Muse Spark 1.2 (xhigh) |
|---|---|---|---|
| Coding | |||
| Arena Coding | Arena (LMArena) | 1511† | 1541 |
| Arena WebDev | Arena (LMArena) | 1401† | 1533 |
| LiveBench 2.0 Coding | LiveBench | – | 78.2%† |
| LiveBench 1.0 Coding | LiveBench | 77.5% | 77.5% |
| SciCode (AA) | Artificial Analysis | – | 57.4% |
| LiveCodeBench | Vals AI | 84.1% | – |
| IOI | Vals AI | – | 21.8% |
| SWE-bench (Vals) | Vals AI | 78.2% | 86.6% |
| SWE-Bench Pro | Scale AI SEAL | 59.1% | – |
| Code Migration | Vals AI | 35.0% | 29.9% |
| ProgramBench | Vals AI | 0.0% | 0.5% |
| Vibe Code Bench v1.1 | Vals AI | 67.4% | 79.1% |
| SWE-Bench Pro (private) | Scale AI SEAL | 59.1% | – |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 40.8% | – |
| SWE Atlas: Refactoring | Scale AI SEAL | 44.3% | – |
| SWE Atlas: Test Writing | Scale AI SEAL | 44.4% | – |
| MirrorCode | Epoch AI Benchmarking Hub | 15.6%† | – |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 51.8% | 54.9% |
| ALE-Bench | Epoch AI Benchmarking Hub | 1607† | – |
| AlgoTune | AlgoTune | 1.9† | – |
| FrontierSWE | Epoch AI Benchmarking Hub | – | 12.0% |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | – | 7.1% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 78.3% | 80.2% |
| Startrise Frontend | Startrise | – | 70.3%† |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 81.8%† | – |
| APEX-Agents | Mercor | 52.4%† | 45.4%† |
| Arena Agent | Arena (LMArena) | – | -7 |
| LiveBench Long Agentic | LiveBench | – | 21.3%† |
| LiveBench Browser | LiveBench | – | 36.0%† |
| LiveBench Agentic Coding | LiveBench | 53.8% | 57.6% |
| Terminal-Bench Hard | Artificial Analysis | 57.6% | – |
| τ²-Bench Telecom (AA) | Artificial Analysis | 87.1% | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | – | 69.7% |
| MCP Atlas | Scale AI SEAL | 70.6% | – |
| HiL-Bench | Scale AI SEAL | 9.7%† | – |
| τ³-bench Banking | τ²-bench | 39.4% | – |
| τ³-Banking (AA) | Artificial Analysis | 39.6% | 34.9% |
| APEX-Agents (AA) | Artificial Analysis | 33.3% | – |
| CyberBench | Vals AI | – | 69.5% |
| Finance Agent v2 | Vals AI | – | 60.6% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 0.0% | 25.4% |
| Legal Research Bench | Vals AI | – | 43.8% |
| SkillsBench | Vals AI | 51.7% | 53.0% |
| SREBench | Vals AI | – | 0.0% |
| Tax Agent Bench | Vals AI | – | 32.8% |
| Terminal-Bench 2.0 (Vals) | Vals AI | 58.4%† | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | – | 6.1% |
| DeepResearch Bench | DeepResearch Bench | 35.1%† | – |
| PostTrainBench | Epoch AI Benchmarking Hub | 19.0%† | – |
| METR Time Horizons | METR | 74.3% | – |
| GBAEval | Epoch AI Benchmarking Hub | 45.1%† | – |
| Vending-Bench 2 | Andon Labs | 6144.2† | – |
| AutomationBench | Artificial Analysis | – | 40.6% |
| GDP.pdf (AA) | Artificial Analysis | – | 17.4% |
| EnterpriseOps-Gym | Artificial Analysis | – | 47.3% |
| AA-Briefcase v1.1 | Artificial Analysis | – | 1327 |
| DeepSearchQA | 63.7%† | – | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 78.6% | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 49.0% | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 95.3% | – |
| ProofBench | Vals AI | 56.0% | 43.0%† |
| LiveBench Mathematics | LiveBench | 94.2% | 91.2% |
| AIME (Vals) | Vals AI | 96.7% | – |
| AIME 2026 | MathArena | 99.2% | – |
| HMMT February 2026 | MathArena | 97.7% | – |
| USAMO 2026 | MathArena | 95.2% | – |
| MathArena Apex | MathArena | 54.2% | – |
| Arena Maths | Arena (LMArena) | 1460† | 1491 |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 45.1% | 60.3% |
| LiveBench 2.0 Data Analysis | LiveBench | – | 76.5%† |
| LiveBench Research | LiveBench | – | 67.7%† |
| LiveBench 1.0 Data Analysis | LiveBench | 79.3% | 76.5% |
| AA-Omniscience | Artificial Analysis | 5.8 | 27.2 |
| MMLU-Pro | Vals AI | 87.5% | 88.3% |
| LegalBench | Vals AI | 86.0% | 85.3% |
| CorpFin | Vals AI | 65.3% | 70.9% |
| TaxEval | Vals AI | 74.0% | 80.4% |
| MedQA | Vals AI | 96.1% | – |
| PRBench Finance | Scale AI SEAL | 45.6%† | – |
| PRBench Legal | Scale AI SEAL | 44.4%† | – |
| MultiNRC | Scale AI SEAL | 58.3%† | – |
| BioMysteryBench | Vals AI | – | 64.8% |
| CaseLaw v2 | Vals AI | 63.8% | – |
| Excel Modeling Benchmark | Vals AI | – | 57.0% |
| MedCode | Vals AI | 41.3% | 49.4% |
| MedScribe | Vals AI | 77.5% | 90.1% |
| Public Benefits Bench v1.1 | Vals AI | – | 68.5% |
| EBR-bench | Epoch AI Benchmarking Hub | 25.4% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | – | 16.0% |
| AA-Omniscience: accuracy | Artificial Analysis | 50.9% | 45.4% |
| AA-Omniscience: non-hallucination | Artificial Analysis | 8.3% | 66.7% |
| Instruction following | |||
| LiveBench Language | LiveBench | 82.6% | 78.6% |
| LiveBench Instruction Following | LiveBench | 70.2% | 74.3% |
| IFBench | Artificial Analysis | 74.0% | – |
| Arena Instruction Following | Arena (LMArena) | 1460† | 1472 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1465† | 1491 |
| EQ-Bench 4 | EQ-Bench | 1260† | – |
| Arena Creative Writing | Arena (LMArena) | 1435† | 1439 |
| Arena Multi-turn | Arena (LMArena) | 1480† | 1511 |
| Arena Longer Queries | Arena (LMArena) | 1477† | 1488 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1292† | 1304 |
| LiveBench Design | LiveBench | – | 67.1%† |
| MMMU-Pro | Artificial Analysis | 78.4% | – |
| VISTA | Scale AI SEAL | 50.9%† | – |
| MMMU-Pro (Vals) | Vals AI | 87.5% | 86.1% |
| MortgageTax | Vals AI | 68.3% | 65.4% |
| SAGE | Vals AI | 43.3% | 47.7% |
| VTB | Scale AI SEAL | 29.2%† | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 27.1%† | – |
| Long context | |||
| AA-LCR | Artificial Analysis | 82.0% | 79.0% |
| Arena Document | Arena (LMArena) | 1470† | – |
| MLCR | Artificial Analysis | – | 7.2% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 157.3† | 155.1† |
| LiveBench 2.0 | LiveBench | – | 62.4%† |
| LiveBench 1.0 | LiveBench | 78.0% | 78.0% |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 39.0 | 39.6 |
| Vals RSI Index | Vals AI | 15.5% | – |
| Vals Multimodal Index | Vals AI | – | 69.8%† |
| Safety & honesty | |||
| MASK | Scale AI SEAL | 89.7%† | – |
Data as of 2026-10-25. Best configuration of each model; every score links to its source on the model pages.
What changed
- Muse Spark 1.2 (xhigh): 1 new result (AA-Briefcase v1.1).
- GPT-5.4 (xhigh): 2 new results (Harvey's Legal Agent Benchmark v1, Vals RSI Index).
- Muse Spark 1.2 (xhigh): 1 new result (Harvey's Legal Agent Benchmark v1).
- Muse Spark 1.2 (xhigh): 2 score revisions.
- Muse Spark 1.2 (xhigh): BenchLeader Index 62.6 → 63.3.
- GPT-5.4 (xhigh): 2 new results (USAMO 2026, τ³-bench Banking).
- Muse Spark 1.2 (xhigh): 4 new results (AutomationBench, GDP.pdf (AA), MLCR…).
- Muse Spark 1.2 (xhigh): 3 new results (Chess Puzzles, ProgramBench, SREBench); 1 score revision.
- GPT-5.4 (xhigh): 1 score revision.
- Muse Spark 1.2 (xhigh): 3 score revisions.
Related comparisons
- GPT-5.4 vs Claude Opus 5.5
- GPT-5.4 vs Gemini 4 Argon
- GPT-5.4 vs Muse Wonder Full
- Muse Spark 1.2 vs Claude Opus 5.5
- Muse Spark 1.2 vs Gemini 4 Argon
- Muse Spark 1.2 vs Muse Wonder Full
- Muse Spark 1.2 vs Claude Fable 5.1
- GPT-5.4 Pro vs Muse Spark 1.2
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
GPT-5.4 vs Muse Spark 1.2: questions
- Is GPT-5.4 better than Muse Spark 1.2?
- GPT-5.4 (xhigh) and Muse Spark 1.2 (xhigh) are neck and neck on quality (64.7 vs 63.2, within the ±5.3 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 2026-10-25, so check the category scores for your use.
- Which is cheaper, GPT-5.4 or Muse Spark 1.2, and will that change?
- Muse Spark 1.2 is cheaper today: $2.00 against $5.63 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is GPT-5.4 better than Muse Spark 1.2 for coding?
- GPT-5.4 scores higher in coding (65 vs 55 on the category index, where 50 is average). On SciCode, the top shared coding benchmark, it is 56.6% for GPT-5.4 against 57.4% for Muse Spark 1.2.
- Is GPT-5.4 better than Muse Spark 1.2 for agents and tool use?
- GPT-5.4 scores higher in agentic tasks (64 vs 55 on the category index, where 50 is average). On τ³-Banking (AA), the top shared agents & tools benchmark, it is 39.6% for GPT-5.4 against 34.9% for Muse Spark 1.2.
- Which is better for long documents?
- GPT-5.4 scores higher on long-context benchmarks (66 vs 65). Both take about 1.1M tokens: GPT-5.4 1,050,000, Muse Spark 1.2 1,048,576.
- Which has lower latency?
- Muse Spark 1.2 (xhigh) answers first: 17.6 s against 118.0 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.
- Which is faster, GPT-5.4 or Muse Spark 1.2?
- Muse Spark 1.2 streams faster: 307 against 96 output tokens per second.
- Do independent benchmarks agree on GPT-5.4 vs Muse Spark 1.2?
- Across the 39 benchmarks both have, GPT-5.4 leads 16, Muse Spark 1.2 leads 22 and 1 is a tie. Every overall index that rates both puts Muse Spark 1.2 ahead.
- When was Muse Spark 1.2 released?
- Muse Spark 1.2 was released on 5 Aug 2026.