GPT-5.4 Nano vs Muse Spark Open
Full profiles: GPT-5.4 Nano (OpenAI) and Muse Spark Open (Meta): every benchmark, price and rank.
Updated · Data from 11 sources
As of 16 Nov 2026, GPT-5.4 Nano and Muse Spark Open are neck and neck on quality (53.9 vs 53.5 on the BenchLeader Index, inside the ±10.9 margin), and GPT-5.4 Nano is 4.3× cheaper at $0.463 vs $2.00 per 1M tokens blended. Choose GPT-5.4 Nano for coding (49 vs 36); choose Muse Spark Open for fast first answers (9.9 s vs 61.3 s) and multimodal (63 vs 47) and reasoning (51 vs 38).
Key facts (as of 16 Nov 2026)
- GPT-5.4 Nano scores 53.9 on the BenchLeader Index and Muse Spark Open scores 53.5 (BenchLeader, 16 Nov 2026).
- GPT-5.4 Nano costs $0.463 per 1M tokens blended (list price, 16 Nov 2026).
- Muse Spark Open costs $2.00 per 1M tokens blended (list price, 16 Nov 2026).
- Of the 16 benchmarks both models have, Muse Spark Open leads 15 (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 | GPT-5.4 Nano | 49 vs 36 category score; SciCode (AA) 47.2% vs 58.5% | High |
| Reasoning | Muse Spark Open | 51 vs 38 category score; ARC-AGI-2 57.5% vs 5.7% | High |
| Long documents | Either | 63 vs 64 category score; AA-LCR v1.1 76.7% vs 78.3% | Either |
| Images and other media | Muse Spark Open | 63 vs 47 category score; MMMU-Pro 80.9% vs 65.4% | High |
| Lowest cost | GPT-5.4 Nano | $0.463 vs $2.00 per 1M blended; agentic coding session $0.017 vs $0.092 | – |
| Fastest first answer | Muse Spark Open | 9.9 s (Muse Spark Open (xhigh)) vs 61.3 s (GPT-5.4 Nano (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 Nano (xhigh) and Muse Spark Open (xhigh) are neck and neck on quality (53.9 vs 53.5, within the ±10.9 margin).
- GPT-5.4 Nano (xhigh) is stronger in coding.
- Muse Spark Open (xhigh) is stronger in long context, multimodal, reasoning.
- GPT-5.4 Nano (xhigh) is 4.3× cheaper ($0.463 vs $2.00 per 1M blended).
At a glance
| Spec | GPT-5.4 Nano | Muse Spark Open |
|---|---|---|
| Developer | OpenAI | Meta |
| Released | 17 Mar 2026 | 15 Oct 2026 |
| Context window (tokens) | 400,000 | 1,048,576 |
| Compared at | GPT-5.4 Nano (xhigh) | Muse Spark Open (xhigh) |
| Open weights | No | Yes |
| Price in / out / cached input per 1M | $0.200 / $1.25 / $0.020 | $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 Nano (xhigh) | Muse Spark Open (xhigh) | Leader |
|---|---|---|---|
| Reasoning | 37.9 | 51.3 | Muse Spark Open |
| Coding | 49.4 | 36.0 | GPT-5.4 Nano |
| Agents & tools | 60.8 | – | not comparable |
| Knowledge | – | 58.8 | not comparable |
| Instruction following | 74.1 | – | not comparable |
| Multimodal | 47.4 | 63.3 | Muse Spark Open |
| Long context | 62.9 | 63.8 | level |
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: Agents & tools, Maths, Knowledge, Instruction following, Human preference.
Where sources disagree
Across the 16 benchmarks both have, GPT-5.4 Nano leads 1, Muse Spark Open leads 15.
| Overall index | GPT-5.4 Nano (xhigh) | Muse Spark Open (xhigh) | Leader |
|---|---|---|---|
| BenchLeader Index | 53.9 | 53.5 | neck and neck |
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 | GPT-5.4 Nano (xhigh) | Muse Spark Open (xhigh) |
|---|---|---|
| Blended price per 1M tokens | $0.463 | $2.00 |
| Input / output per 1M | $0.200 / $1.25 | $1.25 / $4.25 |
| Cached input per 1M | $0.020 | $0.150 |
| Price per task: chat reply | $0.0005 | $0.0018 |
| Price per task: summarize a 30-page report | $0.0032 | $0.018 |
| Price per task: code edit | $0.0031 | $0.014 |
| Price per task: agentic coding session | $0.017 | $0.092 |
| Price per task: structured extraction | $0.0007 | $0.0034 |
| Agentic coding session, input mostly cached | $0.0089 | $0.043 |
| Cost per benchmark task (AA Index) | – | $2.02 |
| Output tokens per task (AA Index) | – | 66k |
| Of which reasoning | – | 32k |
| Time per task (AA Index) | – | 6 min |
| Output speed | 145 tok/s | 180 tok/s |
| Time to first answer | 61.3 s | 9.9 s |
| Time to a complete answer, last measured 13 Nov 2026 | 185.1 s | 39.9 s |
Each model is compared at its best-ranked reasoning-effort setting. GPT-5.4 Nano (xhigh) scores 53.9 and takes 61 s to its first answer token; GPT-5.4 Nano (medium) scores 51.4 in 6 s, GPT-5.4 Nano (high) scores 48.3 in 1 s. Muse Spark Open (xhigh) scores 53.5 and takes 10 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
130 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 Nano (xhigh) | Muse Spark Open (xhigh) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 53.9 | 53.5 |
| Agents & tools score | BenchLeader | 60.8 | – |
| Coding score | BenchLeader | 49.4 | 36.0 |
| Instruction following score | BenchLeader | 74.1 | – |
| Long context score | BenchLeader | 62.9 | 63.8 |
| Multimodal score | BenchLeader | 47.4 | 63.3 |
| Reasoning score | BenchLeader | 37.9 | 51.3 |
| Knowledge score | BenchLeader | – | 58.8 |
| Blended price $/M | BenchLeader | $0.463 | $2.00 |
| Output speed | BenchLeader | 145 tok/s | 180 tok/s |
| Time to first answer | BenchLeader | 61.3 s | 9.9 s |
| Context window | BenchLeader | 400,000 tokens | 1,048,576 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 78.5%† | – |
| Arena Hard Prompts | Arena (LMArena) | 1421† | – |
| LiveBench 2.0 Reasoning | LiveBench | – | 73.7%† |
| LiveBench 1.0 Reasoning | LiveBench | 81.1% | – |
| GPQA Diamond (AA) | Artificial Analysis | 81.7% | 89.0% |
| Humanity's Last Exam (AA) | Artificial Analysis | 28.3% | 43.9% |
| HLE-Diamond (AA) | Artificial Analysis | – | 26.3% |
| GPQA Diamond (Vals) | Vals AI | 77.5%† | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 39.7%† | – |
| ARC-AGI-1 | ARC Prize | 51.5% | 89.0% |
| ARC-AGI-2 | ARC Prize | 5.7% | 57.5% |
| ARC-AGI-3 | ARC Prize | – | 0.4% |
| CritPt | Artificial Analysis | 9.3% | 14.9% |
| CritPt v2 hard subset (AA) | Artificial Analysis | – | 29.5% |
| HLE-Diamond (Scale) | Scale AI SEAL | – | 28.0%† |
| Arena Expert | Arena (LMArena) | 1436† | – |
| Chess Puzzles | Epoch AI Benchmarking Hub | 45.0% | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 6.0%† | – |
| LMCA | Epoch AI Benchmarking Hub | 36.9% | – |
| DTBench | Epoch AI Benchmarking Hub | 80.3% | – |
| ForecastBench | ForecastBench | 58.6%† | – |
| Coding | |||
| SciCode | SciCode | 47.2% | – |
| WeirdML | WeirdML | 49.2%† | – |
| Arena Coding | Arena (LMArena) | 1460† | – |
| Arena WebDev | Arena (LMArena) | – | 1526† |
Show all 130 benchmarks
| Metric | Publisher | GPT-5.4 Nano (xhigh) | Muse Spark Open (xhigh) |
|---|---|---|---|
| Coding | |||
| LiveBench 2.0 Coding | LiveBench | – | 55.1%† |
| LiveBench 1.0 Coding | LiveBench | 70.8% | – |
| SciCode (AA) | Artificial Analysis | 47.2% | 58.5% |
| LiveCodeBench | Vals AI | 84.0%† | – |
| SWE-bench (Vals) | Vals AI | 69.8%† | – |
| Code Migration | Vals AI | 14.5%† | 21.2% |
| Code Migration v1.1 | Vals AI | – | 26.9% |
| Vibe Code Bench v1.1 | Vals AI | 26.1%† | 76.2% |
| SWE Atlas v1.1 | Scale AI SEAL | – | 42.3%† |
| SWE Atlas v1.1: Codebase QnA | Scale AI SEAL | – | 50.0%† |
| SWE Atlas v1.1: Test Writing | Scale AI SEAL | – | 41.1%† |
| SWE Atlas v1.1: Refactoring | Scale AI SEAL | – | 35.7%† |
| SWE-Bench Pro V2 | Scale AI SEAL | – | 95.0% |
| SWE-Bench Pro V2 HARD | Scale AI SEAL | – | 82.3% |
| ALE-Bench | Epoch AI Benchmarking Hub | 1004.5† | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 0.5% | 8.1% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 60.7% | 76.4% |
| GameBench 2 Overall | DuelLab GameBench 2 | – | 55.3† |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 31.1† | 59.3 |
| Design Arena Website | Design Arena | – | 1305 |
| FrontierCode 1.1 (AA) | Artificial Analysis | – | 27.6% |
| FrontierCode 2.0 (AA) | Artificial Analysis | – | 7.6% |
| Agents & tools | |||
| LiveBench Long Agentic | LiveBench | – | 12.1%† |
| LiveBench Browser | LiveBench | – | 26.2%† |
| LiveBench Agentic Coding | LiveBench | 46.8% | – |
| Terminal-Bench Hard | Artificial Analysis | 42.4% | – |
| τ²-Bench Telecom (AA) | Artificial Analysis | 76.0% | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | 41.6%† | – |
| HiL-Bench | Scale AI SEAL | – | 59.7%† |
| GDPval-AA v2.1, normalised score | Artificial Analysis | – | 45.1% |
| GDPval-AA v2.1 | Artificial Analysis | 954 | 1401 |
| τ³-Banking (AA) | Artificial Analysis | 27.4% | – |
| ITBench SRE (AA) | Artificial Analysis | 24.4% | – |
| APEX-Agents (AA) | Artificial Analysis | 24.9% | – |
| Finance Agent v2 | Vals AI | 38.2%† | 61.3% |
| Finance Agent v2.1 | Vals AI | – | 60.7% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | – | 25.0% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 0.0%† | 16.7% |
| Legal Research Bench | Vals AI | 6.3%† | 38.5% |
| Tax Agent Bench | Vals AI | 5.0%† | 26.3% |
| Terminal-Bench 2.0 (Vals) | Vals AI | 39.9%† | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | – | 6.6% |
| AutomationBench | Artificial Analysis | 6.0% | 49.7% |
| GDP.pdf (AA) | Artificial Analysis | 7.8% | 18.6% |
| EnterpriseOps-Gym | Artificial Analysis | 31.5% | – |
| AA-Briefcase v1.1 | Artificial Analysis | 674 | 1280 |
| GDPx | GDPx by SemiAnalysis | – | 1.7%† |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 1.9% |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 44.9%† | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 12.2%† | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 87.8%† | – |
| ProofBench | Vals AI | 5.0%† | – |
| LiveBench Mathematics | LiveBench | 91.0% | – |
| AIME (Vals) | Vals AI | 88.8%† | – |
| ArXivMath | MathArena | – | 27.0% |
| BrokenArXiv | MathArena | – | 30.1% |
| Arena Maths | Arena (LMArena) | 1424† | – |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 11.7%† | – |
| LiveBench 2.0 Data Analysis | LiveBench | – | 60.2%† |
| LiveBench Research | LiveBench | – | 82.7%† |
| LiveBench 1.0 Data Analysis | LiveBench | 67.6% | – |
| AA-Omniscience v1.1 | Artificial Analysis | – | 25.8 |
| MMLU-Pro | Vals AI | 77.2%† | – |
| LegalBench | Vals AI | 77.9%† | – |
| CorpFin | Vals AI | 61.2%† | – |
| TaxEval | Vals AI | 67.4%† | – |
| PRBench Legal | Scale AI SEAL | – | 59.8%† |
| CaseLaw v2 | Vals AI | 51.9%† | – |
| Excel Modeling Benchmark | Vals AI | 44.8%† | 53.7% |
| Excel Modeling Benchmark v1.1 | Vals AI | – | 62.0% |
| MedCode | Vals AI | 41.0%† | – |
| MedScribe | Vals AI | 77.1%† | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | – | 42.0% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | – | 72.0% |
| Instruction following | |||
| LiveBench Language | LiveBench | 62.5% | – |
| LiveBench Instruction Following | LiveBench | 67.2% | – |
| IFBench | Artificial Analysis | 75.9% | – |
| MultiChallenge | Scale AI SEAL | – | 74.2%† |
| Arena Instruction Following | Arena (LMArena) | 1386† | – |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1401† | – |
| Arena Creative Writing | Arena (LMArena) | 1337† | – |
| Arena Multi-turn | Arena (LMArena) | 1414† | – |
| Arena Longer Queries | Arena (LMArena) | 1403† | – |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1196† | – |
| LiveBench Design | LiveBench | – | 58.1%† |
| MMMU-Pro | Artificial Analysis | 65.4% | 80.9% |
| MMMU-Pro (Vals) | Vals AI | 73.6%† | – |
| MortgageTax | Vals AI | 59.1%† | – |
| SAGE | Vals AI | – | 85.8% |
| Humanity's Sixth Sense | Scale AI SEAL | – | 29.6%† |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 76.7% | 78.3% |
| MLCR | Artificial Analysis | 16.1% | 7.8% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 146.4† | 155.4† |
| LiveBench 2.0 | LiveBench | – | 48.5%† |
| LiveBench 1.0 | LiveBench | 69.6% | – |
| AA Intelligence Index v5.0.1 | Artificial Analysis | – | 24.9 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 20.7 | 38.9 |
| AA Progress Index v1 | Artificial Analysis | 270.1 | 306.7 |
| Vals Index | Vals AI | – | 48.6 |
| Vals Finance Index | Vals AI | – | 53.0% |
| Vals Multimodal Index | Vals AI | 47.6%† | – |
| Vals Legal Index | Vals AI | – | 68.5% |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | – | 7.6%† |
| DistressBench | Scale AI SEAL | – | 87.7%† |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 27.8 |
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
- GPT-5.4 Nano (xhigh): 2 new results (GDPval-AA v2.1, AA-Briefcase v1.1).
- Muse Spark Open (xhigh): 1 score revision.
- Muse Spark Open (xhigh): 1 new result (GameBench 2 by reasoning setting); 1 score revision.
- Muse Spark Open (xhigh): BenchLeader Index 53.3 → 53.8.
- Muse Spark Open (xhigh): 1 new result (ARC-AGI-3).
- GPT-5.4 Nano (xhigh): BenchLeader Index 53.6 → 54.1.
- Muse Spark Open (xhigh): BenchLeader Index 59.5 → 53.3.
- GPT-5.4 Nano (xhigh): 1 new result (AA Progress Index v1).
- Muse Spark Open (xhigh): 6 new results (AA Intelligence Index v5.0.1, AA Progress Index v1, HLE-Diamond (AA)…); 3 score revisions; cost per task moved from $0.84/M to $2.02/M (139% up).
- Muse Spark Open (xhigh): 6 new results (MLCR, Code Migration v1.1, Excel Modeling Benchmark v1.1…); cached input price moved from $1.25/M to $0.15/M (88% down).
Related comparisons
- GPT-5.4 Nano vs MiniMax M3.1 Flash
- Muse Spark Open vs Claude Opus 5.5
- Muse Spark Open vs GPT-6 Astra
- Muse Spark Open vs GPT-6.1 Sol
- Muse Spark Open vs Claude Fable 5
- Muse Spark Open vs Gemini 4 Argon
- Muse Spark Open vs GPT-5.6 Sol
- Muse Spark Open vs Claude Sonnet 5.5
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
GPT-5.4 Nano vs Muse Spark Open: questions
- Is GPT-5.4 Nano better than Muse Spark Open?
- GPT-5.4 Nano (xhigh) and Muse Spark Open (xhigh) are neck and neck on quality (53.9 vs 53.5, within the ±10.9 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
- GPT-5.4 Nano vs Muse Spark Open: which is cheaper, and will that change?
- GPT-5.4 Nano is cheaper today: $0.463 against $2.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is GPT-5.4 Nano or Muse Spark Open better for coding?
- GPT-5.4 Nano scores higher in coding (49 vs 36 on the category index, where 50 is average).
- Which is faster, GPT-5.4 Nano or Muse Spark Open?
- Muse Spark Open streams faster (180 vs 145 output tokens per second, Artificial Analysis 7-day median).
- Which is better for long documents?
- They are level on long-context benchmarks (63 vs 64). Muse Spark Open accepts more context: 1,048,576 against 400,000 tokens for GPT-5.4 Nano.
More questions (3)
- Which has lower latency?
- Muse Spark Open answers first: 9.9 s against 61.3 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 GPT-5.4 Nano vs Muse Spark Open?
- Across the 16 benchmarks both have, GPT-5.4 Nano leads 1, Muse Spark Open leads 15.
- When was Muse Spark Open released?
- Muse Spark Open was released on 15 Oct 2026.