MetaOpen weights: meta-models/Muse-Spark-Open-816B-A42BReasoning modelNewReleased 15 Oct 2026Compare with another modelMuse Spark Open
Muse Spark Open is a Meta open-weights reasoning model, released 15 Oct 2026. Its best configuration (xhigh reasoning effort) ranks #85 of 761 on the BenchLeader Index at 61.5 ±3.7, in the upper half. The ± is the point: 174 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in composite (74) and lowest in long context (65). At $2.00 per million tokens blended it is pricier than most ranked models. Output speed of 131 tokens per second puts it faster than most, with a first answer in 16.0 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 15 Oct 2026.
- Blended price
- $2.00/M
- $1.25 in · $4.25 out
- Output speed
- 131 tok/s
- measured by Artificial Analysis
- First answer
- 16 s
- first token 0.41 s
- Context
- 1.0M
- Full answer
- 20 s
- median, reasoning included
- Cost per run
- $0.844
- one full Intelligence Index run
- Released
- 15 Oct 2026
- Overall index62
- Reasoning66
- Knowledge73
- Multimodal65
- Long context65
- Composite74
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Metadata
| Benchmark | xhigh | Source | Trend |
|---|---|---|---|
| AA Openness Indexnot in index | 27.8#35 | Artificial Analysis |
Reasoning
| Benchmark | xhigh | Source | Trend |
|---|---|---|---|
| GPQA Diamond (AA)not in index | 89.0%#77 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 43.9%#50 | Artificial Analysis | |
| CritPt | 14.9%#86 | Artificial Analysis |
Coding
| Benchmark | xhigh | max | Source | Trend |
|---|---|---|---|---|
| Arena WebDev | – | 1511#53 | Arena (LMArena) | |
| SciCode (AA)not in index | 58.5%#23 | – | Artificial Analysis | |
| Code Migrationnot in index | 21.2%#49 | – | Vals AI | |
| Vibe Code Bench v1.1not in index | 76.2%#33 | – | Vals AI | |
| Terminal-Bench 4.0 (AA)not in index | 8.1%#81 | – | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | 76.4%#56 | – | Artificial Analysis |
Agents & tools
| Benchmark | xhigh | Source | Trend |
|---|---|---|---|
| GDPval-AA v2.1not in index | 45.1%#61 | Artificial Analysis | |
| Finance Agent v2not in index | 61.3%#3 | Vals AI | |
| Harvey's Legal Agent Benchmarknot in index | 16.7%#9 | Vals AI | |
| Legal Research Benchnot in index | 38.5%#32 | Vals AI | |
| Tax Agent Benchnot in index | 26.3%#33 | Vals AI | |
| Terminal-Bench 4.0 (Vals)not in index | 6.6%#38 | Vals AI | |
| AutomationBenchnot in index | 49.7%#41 | Artificial Analysis | |
| GDP.pdfnot in index | 18.6%#35 | Artificial Analysis | |
| AA-Briefcase v1.1not in index | 1279#44 | Artificial Analysis |
Knowledge
| Benchmark | xhigh | Source | Trend |
|---|---|---|---|
| AA-Omniscience | 24.6#45 | Artificial Analysis | |
| Excel Modeling Benchmarknot in index | 53.7%#48 | Vals AI | |
| AA-Omniscience: accuracynot in index | 39.3%#96 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 75.9%#19 | Artificial Analysis |
Multimodal
| Benchmark | xhigh | Source | Trend |
|---|---|---|---|
| MMMU-Pro | 80.9%#37 | Artificial Analysis |
Long context
| Benchmark | xhigh | Source | Trend |
|---|---|---|---|
| AA-LCR | 78.3%#100 | Artificial Analysis |
Composite
| Benchmark | xhigh | Source | Trend |
|---|---|---|---|
| AA Intelligence Index v4.3.2 | 38.9#65 | Artificial Analysis | |
| Vals Indexnot in index | 46.6#37 | Vals AI |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Baseten | 196 tok/s | 0.28 s | $1.25 | $4.25 | 1.0M | fp4 |
| Nebius Token Factory | 105 tok/s | 0.54 s | $1.25 | $4.25 | 1.0M | fp4 |
| Fireworks | 82 tok/s | 0.66 s | $1.25 | $4.25 | 1.0M | – |
| Modal | 79 tok/s | 7.58 s | $1.25 | $4.25 | 1M | – |
| Together | 77 tok/s | 0.38 s | $1.25 | $4.25 | 1.0M | – |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $1.25 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0018 | $0.0018 | 18.3 s |
| Summarise a 30-page report | 12,000 / 600 | $0.018 | $0.018 | 20.6 s |
| Code edit | 6,000 / 1,500 | $0.014 | $0.014 | 27.5 s |
| Agentic coding session | 60,000 / 4,000 | $0.092 | $0.092 | 46.6 s |
| Structured extraction | 2,000 / 200 | $0.0034 | $0.0034 | 17.6 s |
More from Meta
Compare Muse Spark Open with…
- Muse Spark Open vs Claude Opus 5.5
- Muse Spark Open vs Gemini 4 Argon
- Muse Spark Open vs Claude Fable 5.1
- 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 Muse Wonder Full
- Muse Spark Open vs Claude Opus 5
- Muse Spark Open vs GPT-5.6 Sol
- Muse Spark Open vs Claude Sonnet 5.5
See also
Data as of 18 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Muse Spark Open: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/muse-spark-open, data as of 18 Oct 2026.