Model latency
Time to first token and output throughput from the last 24 hours, grouped by Freebuff model.
Updated 28 min ago
Median TTFT
2.3s
7.6M samples in 24h
P95 TTFT
7.0s
Slower tail across all models
Output speed
91 tok/s
Output tokens per second after TTFT, stalls and bursts left out
Lowest Median
858ms
Nemotron 3.5 Lightning · 260K samples
Models
13
Reporting TTFT samples
By model
Sorted by median TTFT. Bars show P95 latency.
Last 24 hours
Nemotron 3.5 Lightning
Median
858ms
P95
2.1s
Tok/s
170
260K sampleshourly median
Beam 1.0
Median
910ms
P95
1.6s
Tok/s
172
260K sampleshourly median
Claude Haiku 5.5 · 100k context
Median
1.3s
P95
6.2s
Tok/s
334
55K sampleshourly median
DeepSeek V4.1 Flash Fast
Median
2.2s
P95
7.0s
Tok/s
111
3.5M sampleshourly median
Muse Ember
Median
2.2s
P95
4.1s
Tok/s
681
16K sampleshourly median
DeepSeek V4.1 Flash
Median
2.3s
P95
6.0s
Tok/s
100
1.3M sampleshourly median
DeepSeek V4.1 Flash Slow
Median
2.3s
P95
6.4s
Tok/s
40
1.5M sampleshourly median
Solar Pro 4
Median
2.8s
P95
8.9s
Tok/s
91
310K sampleshourly median
Solar Mini 4
Median
3.0s
P95
6.2s
Tok/s
100
130K sampleshourly median
Muse Spark 1.3
Median
3.2s
P95
8.4s
Tok/s
134
130K sampleshourly median
GPT-6.1 Terra
Median
4.2s
P95
12s
Tok/s
186
14K sampleshourly median
Glyph Cluster
Median
5.9s
P95
17s
Tok/s
72
40K sampleshourly median
GPT-6 Luna
Median
6.0s
P95
23s
Tok/s
267
61K sampleshourly median