Live Operations
Current network state — inference load, node capacity, quality pulse.
Live
Requests / min+6.4%
552
Derived from 30d throughput with current-load uplift
Active jobs+4.1%
1,703
Tokens / sec+3.2%
292K
Across 4,399 processing GPUs
p50 latency-1.4%
483ms
Error rate-0.4%
3.2%
Active jobs by model (now)
Estimated from 30d request share × 1,703 processing nodes
Llama 3.1 70B InstructA
373 jobs · 21.9%Llama 3.1 405B InstructS
251 jobs · 14.7%DeepSeek-V3A
243 jobs · 14.3%DeepSeek Coder V2A
185 jobs · 10.9%DeepSeek-R1A
152 jobs · 8.9%Qwen 2 VL 72BAerr 3.3%
125 jobs · 7.4%Qwen 2.5 72B InstructA
117 jobs · 6.9%Mistral Large 2Aerr 3.3%
104 jobs · 6.1%QwQ 32B PreviewBerr 3.1%
86 jobs · 5%Mixtral 8x22B InstructAerr 3.4%
68 jobs · 4%Node capacity (now)
Processing1,703 (37%)
Idle — available666 (14%)
Offline997 (22%)
Degraded181 (4%)
Total: 4,603Active: 3,164
Quality pulse
Completion rate97.8%
Median RS71
Avg uptime83%
Dev error rate3.2%
p50 latency483ms
Supply by region
Active nodes and active developers across top regions
EU1,086 nodes · 1,093 devs
Asia796 nodes · 894 devs
US454 nodes · 480 devs
MENA386 nodes · 384 devs
LATAM245 nodes · 296 devs
Oceania100 nodes · 119 devs