DeepSeek V4 Flash vs Muse Spark 1.3
Full profiles: DeepSeek V4 Flash (DeepSeek) and Muse Spark 1.3 (Meta): every benchmark, price and rank.
Updated · Data from 10 sources
As of 16 Nov 2026, DeepSeek V4 Flash and Muse Spark 1.3 are neck and neck on quality (57.2 vs 61.6 on the BenchLeader Index, inside the ±9.1 margin), and DeepSeek V4 Flash is 7.6× cheaper at $0.262 vs $2.00 per 1M tokens blended. Choose DeepSeek V4 Flash for maths (62 vs 56) and fast first answers (0.7 s vs 2.6 s); choose Muse Spark 1.3 for knowledge (81 vs 55) and reasoning (71 vs 59), where it also streams ~1.5× faster.
Key facts (as of 16 Nov 2026)
- DeepSeek V4 Flash scores 57.2 on the BenchLeader Index and Muse Spark 1.3 scores 61.6 (BenchLeader, 16 Nov 2026).
- DeepSeek V4 Flash costs $0.262 per 1M tokens blended (list price, 16 Nov 2026).
- Muse Spark 1.3 costs $2.00 per 1M tokens blended (list price, 16 Nov 2026).
- Muse Spark 1.3 streams ~1.5× faster than DeepSeek V4 Flash (Artificial Analysis, 7-day median, 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 |
|---|---|---|---|
| Maths | DeepSeek V4 Flash | 62 vs 56 category score | Medium |
| Reasoning | Muse Spark 1.3 | 71 vs 59 category score | High |
| Knowledge | Muse Spark 1.3 | 81 vs 55 category score | High |
| Chat quality (human preference) | – | Only DeepSeek V4 Flash has an Arena rating (62 category score) | Not comparable |
| Lowest cost | DeepSeek V4 Flash | $0.262 vs $2.00 per 1M blended; agentic coding session $0.011 vs $0.092 | – |
| Fastest first answer | DeepSeek V4 Flash | 0.7 s (DeepSeek V4 Flash (high)) vs 2.6 s (Muse Spark 1.3) 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
- DeepSeek V4 Flash (high) and Muse Spark 1.3 are neck and neck on quality (57.2 vs 61.6, within the ±9.1 margin).
- DeepSeek V4 Flash (high) is stronger in maths.
- Muse Spark 1.3 is stronger in knowledge, reasoning.
- DeepSeek V4 Flash (high) is 7.6× cheaper ($0.262 vs $2.00 per 1M blended).
- Muse Spark 1.3 streams 2.0× faster (163 vs 82 tokens per second).
At a glance
| Spec | DeepSeek V4 Flash | Muse Spark 1.3 |
|---|---|---|
| Developer | DeepSeek | Meta |
| Released | 31 Jul 2026 | 2 Sep 2026 |
| Context window (tokens) | 1,048,576 | 1,048,576 |
| Compared at | DeepSeek V4 Flash (high) | Muse Spark 1.3 |
| Open weights | Yes | No |
| Price in / out / cached input per 1M | $0.150 / $0.600 / $0.0030 | $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 | DeepSeek V4 Flash (high) | Muse Spark 1.3 | Leader |
|---|---|---|---|
| Reasoning | 59.2 | 71.0 | Muse Spark 1.3 |
| Coding | 55.0 | – | not comparable |
| Agents & tools | – | 61.3 | not comparable |
| Maths | 61.9 | 56.3 | DeepSeek V4 Flash |
| Knowledge | 54.8 | 81.2 | Muse Spark 1.3 |
| Human preference | 62.0 | – | not comparable |
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: Coding, Agents & tools, Instruction following, Long context, Multimodal, Human preference.
Where sources disagree
Across the 0 benchmarks both have, DeepSeek V4 Flash leads 0, Muse Spark 1.3 leads 0. Every overall index that rates both puts Muse Spark 1.3 ahead.
| Overall index | DeepSeek V4 Flash (high) | Muse Spark 1.3 | Leader |
|---|---|---|---|
| BenchLeader Index | 57.2 | 61.6 | neck and neck |
| Arena Text | 1439 | 1497† | Muse Spark 1.3 |
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 | DeepSeek V4 Flash (high) | Muse Spark 1.3 |
|---|---|---|
| Blended price per 1M tokens | $0.262 | $2.00 |
| Input / output per 1M | $0.150 / $0.600 | $1.25 / $4.25 |
| Cached input per 1M | $0.0030 | $0.150 |
| Price per task: chat reply | $0.0002 | $0.0018 |
| Price per task: summarize a 30-page report | $0.0022 | $0.018 |
| Price per task: code edit | $0.0018 | $0.014 |
| Price per task: agentic coding session | $0.011 | $0.092 |
| Price per task: structured extraction | $0.0004 | $0.0034 |
| Agentic coding session, input mostly cached | $0.0048 | $0.043 |
| Output speed | 82 tok/s | 163 tok/s |
| Time to first answer | 0.7 s | 2.6 s |
Each model is compared at its best-ranked reasoning-effort setting. DeepSeek V4 Flash (high) scores 57.2 and takes 1 s to its first answer token; DeepSeek V4 Flash (max) scores 56.8 in 28 s, DeepSeek V4 Flash scores 53.8 in 1 s. Muse Spark 1.3 scores 61.6 and takes 3 s to its first answer token; Muse Spark 1.3 (max) scores 60.6 in 31 s, Muse Spark 1.3 (xhigh) scores 58.1 in 11 s. 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
143 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 | DeepSeek V4 Flash (high) | Muse Spark 1.3 |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 57.2 | 61.6 |
| Coding score | BenchLeader | 55.0 | – |
| Human preference score | BenchLeader | 62.0 | – |
| Knowledge score | BenchLeader | 54.8 | 81.2 |
| Maths score | BenchLeader | 61.9 | 56.3 |
| Reasoning score | BenchLeader | 59.2 | 71.0 |
| Agents & tools score | BenchLeader | – | 61.3 |
| Blended price $/M | BenchLeader | $0.262 | $2.00 |
| Output speed | BenchLeader | 82 tok/s | 163 tok/s |
| Time to first answer | BenchLeader | 0.7 s | 2.6 s |
| Context window | BenchLeader | 1,048,576 tokens | 1,048,576 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 91.0%† | – |
| SimpleBench | SimpleBench | 61.1%† | 81.8% |
| Arena Hard Prompts | Arena (LMArena) | 1459 | 1517† |
| LiveBench 2.0 Reasoning | LiveBench | – | 89.7%† |
| LiveBench 1.0 Reasoning | LiveBench | 86.6%† | 89.7%† |
| GPQA Diamond (AA) | Artificial Analysis | 90.8%† | 93.5%† |
| Humanity's Last Exam (AA) | Artificial Analysis | 38.5%† | 48.7%† |
| HLE-Diamond (AA) | Artificial Analysis | – | 33.7%† |
| GPQA Diamond (Vals) | Vals AI | 89.9% | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 52.2%† | – |
| ARC-AGI-1 | ARC Prize | 87.0% | – |
| ARC-AGI-2 | ARC Prize | 56.0% | – |
| CritPt | Artificial Analysis | 16.6%† | 24.9%† |
| CritPt v2 hard subset (AA) | Artificial Analysis | – | 48.0%† |
| MysteryMechanism | Vals AI | – | 36.0%† |
| HLE-Diamond (Scale) | Scale AI SEAL | – | 25.4% |
| Arena Expert | Arena (LMArena) | 1467 | 1521† |
| Chess Puzzles | Epoch AI Benchmarking Hub | 32.0%† | 37.0%† |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 34.0%† | 25.0%† |
| LMCA | Epoch AI Benchmarking Hub | 41.7%† | 53.9% |
| DTBench | Epoch AI Benchmarking Hub | 90.9%† | 96.5% |
| ForecastBench | ForecastBench | – | 65.5%† |
| BTF-3 | Epoch AI Benchmarking Hub | – | 14.1%† |
| Coding | |||
| SciCode | SciCode | 42.0% | 58.8%† |
| WeirdML | WeirdML | 57.0% | – |
Show all 143 benchmarks
| Metric | Publisher | DeepSeek V4 Flash (high) | Muse Spark 1.3 |
|---|---|---|---|
| Coding | |||
| FrontierCode | Cognition | 18.8%† | – |
| Arena Coding | Arena (LMArena) | 1480 | 1534† |
| Arena WebDev | Arena (LMArena) | 1582 | 1655† |
| LiveBench 2.0 Coding | LiveBench | – | 84.4%† |
| LiveBench 1.0 Coding | LiveBench | 75.0%† | 81.1%† |
| SciCode (AA) | Artificial Analysis | 50.4%† | 58.8%† |
| LiveCodeBench | Vals AI | 87.3% | – |
| IOI | Vals AI | 32.7% | 56.6%† |
| SWE-bench (Vals) | Vals AI | 88.8% | – |
| Code Migration | Vals AI | 38.6% | 47.4%† |
| ProgramBench | Vals AI | 0.0% | 2.5%† |
| SWE-bench Multimodal | Vals AI | – | 43.1%† |
| Vibe Code Bench 1-100 | Vals AI | – | 20.5%† |
| Vibe Code Bench v1.1 | Vals AI | 74.7% | 85.9%† |
| MirrorCode | Epoch AI Benchmarking Hub | – | 54.4%† |
| CursorBench | Cursor | – | 41.6%† |
| ALE-Bench | Epoch AI Benchmarking Hub | 678.2 | 1688.3† |
| FrontierSWE | Epoch AI Benchmarking Hub | – | 25.8%† |
| WeirdML v3 | WeirdML | – | 6.7%† |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 12.1%† | 33.3%† |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 78.7%† | 84.3%† |
| Startrise Frontend | Startrise | – | 74.7% |
| GameBench 2 Overall | DuelLab GameBench 2 | 34.1† | 57.7 |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 43.4 | 68.5† |
| Design Arena Website | Design Arena | 1199† | 1353† |
| FrontierCode 1.1 (AA) | Artificial Analysis | – | 40.6%† |
| FrontierCode 2.0 (AA) | Artificial Analysis | – | 23.1%† |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | – | 14.8%† |
| APEX-Agents | Mercor | – | 58.6% |
| Arena Agent | Arena (LMArena) | – | 2.2† |
| LiveBench Long Agentic | LiveBench | – | 47.9%† |
| LiveBench Browser | LiveBench | – | 30.9%† |
| LiveBench Agentic Coding | LiveBench | 46.8%† | 64.1%† |
| Terminal-Bench 2.1 (Vals) | Vals AI | 67.0%† | 79.0%† |
| HiL-Bench | Scale AI SEAL | 47.8%† | 66.9% |
| GDPval-AA v2.1 | Artificial Analysis | 1437† | 1681† |
| τ³-Banking (AA) | Artificial Analysis | 39.4%† | 50.5%† |
| ITBench SRE (AA) | Artificial Analysis | – | 33.2%† |
| Analyst Agent (AA) | Artificial Analysis | – | 37.5%† |
| CUA-bench | Vals AI | – | 5.8%† |
| CyberBench | Vals AI | 72.1% | 72.7%† |
| Finance Agent v2 | Vals AI | 49.5% | 60.0%† |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 8.3% | 23.8%† |
| Legal Research Bench | Vals AI | 30.3% | 55.3%† |
| SkillsBench | Vals AI | 50.7% | – |
| SREBench | Vals AI | – | 3.8%† |
| Tax Agent Bench | Vals AI | 28.1% | 42.4%† |
| Terminal-Bench 4.0 (Vals) | Vals AI | 18.7% | 24.8%† |
| Terminal-Bench Science | Vals AI | 4.3% | 10.0%† |
| Arena Agent Task Completion | Arena (LMArena) | – | 3.7† |
| AutomationBench | Artificial Analysis | 54.0%† | 57.9%† |
| GDP.pdf (AA) | Artificial Analysis | 11.0%† | 26.6%† |
| Harvey LAB-AA v1.1 | Harvey | – | 8.9%† |
| AA-Briefcase v1.1 | Artificial Analysis | 1255† | 1581† |
| Design Arena Agentic | Design Arena | – | 1283† |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 10.9%† |
| CWE-Bench (AA) | Artificial Analysis | – | 60.8%† |
| DeepSec-Bench (AA) | Artificial Analysis | – | 12.5%† |
| CyberGym end-to-end (AA) | Artificial Analysis | – | 59.5%† |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 57.5%† | 74.0%† |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 24.4%† | 46.3%† |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 94.4%† | 99.2%† |
| ProofBench | Vals AI | 56.0%† | 55.0% |
| LiveBench Mathematics | LiveBench | 86.8%† | 96.0%† |
| AIME 2026 | MathArena | 95.8%† | – |
| HMMT February 2026 | MathArena | 93.9%† | – |
| MathArena Apex | MathArena | 27.1%† | – |
| Arena Maths | Arena (LMArena) | 1441 | 1506† |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 33.6%† | – |
| LiveBench 2.0 Data Analysis | LiveBench | – | 79.6%† |
| LiveBench Research | LiveBench | – | 82.5%† |
| LiveBench 1.0 Data Analysis | LiveBench | 79.3%† | 79.6%† |
| AA-Omniscience v1.1 | Artificial Analysis | – | 22.4† |
| MMLU-Pro | Vals AI | 86.2% | – |
| LegalBench | Vals AI | 77.7% | – |
| CorpFin | Vals AI | 61.9% | – |
| TaxEval | Vals AI | 70.7% | – |
| PRBench Finance | Scale AI SEAL | – | 59.5% |
| PRBench Legal | Scale AI SEAL | – | 61.6% |
| BioMysteryBench | Vals AI | 64.4% | – |
| Excel Modeling Benchmark | Vals AI | 57.0% | 67.4%† |
| MedCode | Vals AI | 41.4% | – |
| MedScribe | Vals AI | 80.4% | – |
| Access to Justice Benchmark | Vals AI | – | 72.2%† |
| Financial Coaching Bench | Vals AI | – | 70.3%† |
| Immigration Bench | Vals AI | – | 64.9%† |
| DrugDiscoveryBench | Scale AI SEAL | – | 62.2%† |
| GDP.pdf | Epoch AI Benchmarking Hub | – | 27.6% |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | – | 38.0%† |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | – | 74.8%† |
| Instruction following | |||
| LiveBench Language | LiveBench | 79.2%† | 82.8%† |
| LiveBench Instruction Following | LiveBench | 65.5%† | 78.0%† |
| Arena Instruction Following | Arena (LMArena) | 1434 | 1485† |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1439 | 1497† |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1559† | 1907 |
| Arena Creative Writing | Arena (LMArena) | 1411 | 1468† |
| Arena Multi-turn | Arena (LMArena) | 1447 | 1493† |
| Arena Longer Queries | Arena (LMArena) | 1449 | 1498† |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | – | 1310† |
| LiveBench Design | LiveBench | – | 74.8%† |
| MMMU-Pro | Artificial Analysis | – | 82.0%† |
| Humanity's Sixth Sense | Scale AI SEAL | – | 37.4%† |
| GeoBench | GeoBench | – | 86.0% |
| Code2Video-AA | Artificial Analysis | – | 1019† |
| AA-3D | Artificial Analysis | – | 1032† |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 79.7%† | 83.0%† |
| Arena Document | Arena (LMArena) | – | 1474† |
| MLCR | Artificial Analysis | – | 13.9%† |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 154.4† | 158.2 |
| LiveBench 2.0 | LiveBench | – | 68.6%† |
| LiveBench 1.0 | LiveBench | 74.2%† | 81.6%† |
| AA Intelligence Index v5.0.1 | Artificial Analysis | – | 36.6† |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 34.3† | 48.1† |
| AA Progress Index v1 | Artificial Analysis | 309.4† | 324.5† |
| Vals RSI Index | Vals AI | – | 19.6%† |
| AA Cyber Index | Artificial Analysis | – | 44.3† |
| Safety & honesty | |||
| DistressBench | Scale AI SEAL | – | 84.9% |
| Metadata | |||
| AA Openness Index | Artificial Analysis | 44.4† | – |
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
- Muse Spark 1.3: BenchLeader Index 60.2 → 61.6.
- DeepSeek V4 Flash (high): 1 score revision.
- Muse Spark 1.3: 1 score revision.
- Muse Spark 1.3: BenchLeader Index 59.2 → 60.2.
- Muse Spark 1.3: 2 new results (SimpleBench, GeoBench).
- Muse Spark 1.3: BenchLeader Index 57.7 → 59.2.
- Muse Spark 1.3: 4 score revisions.
- Muse Spark 1.3: BenchLeader Index 58.2 → 57.7.
- Muse Spark 1.3: 1 score revision.
- DeepSeek V4 Flash (high): 1 score revision.
Related comparisons
- DeepSeek V4 Flash vs Claude Fable 5.1
- DeepSeek V4 Flash vs GPT-6.1 Sol
- Muse Spark 1.3 vs Claude Fable 5.2
- Muse Spark 1.3 vs Gemini 4 Carbon
- Muse Spark 1.3 vs Claude Opus 5.5
- Muse Spark 1.3 vs GPT-6 Astra
- DeepSeek V4 Flash vs Gemini 4 Argon
- DeepSeek V4 Flash vs Claude Sonnet 5.5
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
DeepSeek V4 Flash vs Muse Spark 1.3: questions
- Is DeepSeek V4 Flash better than Muse Spark 1.3?
- DeepSeek V4 Flash (high) and Muse Spark 1.3 are neck and neck on quality (57.2 vs 61.6, within the ±9.1 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.
- DeepSeek V4 Flash vs Muse Spark 1.3: which is cheaper, and will that change?
- DeepSeek V4 Flash is cheaper today: $0.262 against $2.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Which is faster, DeepSeek V4 Flash or Muse Spark 1.3?
- Muse Spark 1.3 streams ~1.5× faster (163 vs 82 output tokens per second, Artificial Analysis 7-day median).
- Which is better for long documents?
- Both accept 1,048,576 tokens of context.
- Which has lower latency?
- DeepSeek V4 Flash answers first: 0.7 s against 2.6 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.
More questions (1)
- When was Muse Spark 1.3 released?
- Muse Spark 1.3 was released on 2 Sep 2026.