Muse Spark Open
meta/muse-spark-openInputOutput
Compare its 5 providersIncident RSSCompare with Muse Glimmer 30B, Muse Spark 1.1, Muse Spark 1.2
CompareMuse Spark Open
| State | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Healthy | 100% | 100% | 100% | 100% | 527ms | 1.98s | 227 tps | 1.0M | $1.25 / $4.25 | |
Fireworksfireworks | Healthy | 100% | 86.63% | 87.59% | 87.59% | 533ms | 19.9s | 59 tps | 1.0M | $1.25 / $4.25 |
Modalmodal | Healthy | 100% | 100% | 100% | 100% | 2.63s | 74.3s | 38 tps | 1.0M | $1.25 / $4.25 |
| Healthy | 100% | 88.52% | 100% | 100% | 1.20s | 8.68s | 95 tps | 1.0M | $1.25 / $4.25 | |
Togethertogether | Healthy | 100% | 94.52% | 99.35% | 99.35% | 533ms | 33.3s | 73 tps | 1.0M | $1.25 / $4.25 |
- BaseTen100%Healthy24h 100%
- 7d
- 100%
- 30d
- 100%
- p50
- 527ms
- p99
- 1.98s
- Throughput
- 227 tps
- Context
- 1.0M
- $/Mtok
- $1.25 / $4.25
baseten/fp4fp4 - Fireworks100%Healthy24h 86.63%
- 7d
- 87.59%
- 30d
- 87.59%
- p50
- 533ms
- p99
- 19.9s
- Throughput
- 59 tps
- Context
- 1.0M
- $/Mtok
- $1.25 / $4.25
fireworks - Modal100%Healthy24h 100%
- 7d
- 100%
- 30d
- 100%
- p50
- 2.63s
- p99
- 74.3s
- Throughput
- 38 tps
- Context
- 1.0M
- $/Mtok
- $1.25 / $4.25
modal - Nebius100%Healthy24h 88.52%
- 7d
- 100%
- 30d
- 100%
- p50
- 1.20s
- p99
- 8.68s
- Throughput
- 95 tps
- Context
- 1.0M
- $/Mtok
- $1.25 / $4.25
nebius/fp4fp4 - Together100%Healthy24h 94.52%
- 7d
- 99.35%
- 30d
- 99.35%
- p50
- 533ms
- p99
- 33.3s
- Throughput
- 73 tps
- Context
- 1.0M
- $/Mtok
- $1.25 / $4.25
together
Measured 3m ago. Latency and throughput are OpenRouter's rolling 30-minute windows.
History
Availability
Axis starts at 90% so small differences stay visible; it expands to 0 when an endpoint falls below.
Time to first token
Logarithmic axis — the fleet spans two orders of magnitude. P50 is OpenRouter's rolling 30-minute window, averaged over the hour.
Throughput
Median tokens per second.
Daily availability
One row per endpoint · one dot per day · last 90 days.
worst hour ≥ 95%90–95%< 90%no data
Colour by each endpoint's worst hour of each UTC day · 5 endpoints · 1 day with data, 2 endpoint-days below 90%.