Provider health · 8
last 5m · auto-refresh 30sOpenAIconnected
us-east, eu-west
rps
620
p95
1.6s
err%
0.08
4 keys$14,820/d
Anthropicconnected
us-east, eu-west
rps
529
p95
1.6s
err%
0.08
3 keys$15,520/d
Google Vertexconnected
us-central, eu-west4
rps
644
p95
1.2s
err%
0.08
2 keys$7,210/d
xAIdegraded
us-east
rps
18
p95
4.8s
err%
2.40
1 key$980/d
AWS Bedrockconnected
us-east-1, eu-west-1
rps
188
p95
1.8s
err%
0.08
2 keys$2,440/d
Azure OpenAIrotating
eastus2, swedencentral
rps
142
p95
1.4s
err%
0.60
2 keys$3,120/d
Self-hosted (vLLM)connected
k8s · us-east
rps
132
p95
1.8s
err%
0.08
0 keys$412/d
Mistralconnected
eu-west
rps
44
p95
1.4s
err%
0.08
1 key$184/d
Fallback chains
primary → secondary → shadow/v1/chat (default)
primarygpt-5
82%secondaryclaude-4.5-sonnet
14%shadowllama-4-70b
4%/v1/chat (long-context)
primarygemini-3-pro
91%secondaryclaude-4.5-sonnet
7%shadowgpt-5
2%/v1/classify
primarygpt-5-nano
76%secondarygemini-3-flash
22%shadowmistral-large-2
2%Model registry · 11
| Provider | Model | Context | $/1k in | $/1k out | RPS | Status |
|---|---|---|---|---|---|---|
| OpenAI | gpt-5 | 400k | $0.005 | $0.015 | 184 | active |
| OpenAI | gpt-5-mini | 400k | $0.0008 | $0.003 | 412 | active |
| OpenAI | gpt-5-nano | 128k | $0.0002 | $0.0008 | 24 | shadow |
| Anthropic | claude-4.5-sonnet | 500k | $0.003 | $0.015 | 211 | active |
| Anthropic | claude-4-haiku | 200k | $0.00025 | $0.00125 | 318 | active |
| gemini-3-pro | 1M | $0.00125 | $0.005 | 142 | active | |
| gemini-3-flash | 1M | $0.00018 | $0.0007 | 502 | active | |
| xAI | grok-3 | 256k | $0.002 | $0.01 | 18 | shadow |
| Meta (self-host) | llama-4-70b | 128k | $0.00015 | $0.00015 | 88 | active |
| Mistral (self-host) | mistral-large-2 | 128k | $0.0001 | $0.0001 | 44 | active |
| OpenAI | gpt-4o | 128k | $0.0025 | $0.01 | 3 | deprecated |
Routing rules
evaluated top → bottom01
if route = /v1/classify AND tokens < 1k → gpt-5-nano
cheap classification
02
if app = legal-review → claude-4.5-sonnet
long-context, formal tone
03
if tokens_in > 200k → gemini-3-pro
1M ctx window
04
if team = support-ai AND p95 > 3s → claude-4-haiku
latency budget
05
if fallback (primary 5xx) → claude-4.5-sonnet
auto-failover
06
if default → gpt-5
general purpose
Traffic split · /v1/chat
last 1h · 412k requestsgpt-552%
claude-4.5-sonnet31%
gemini-3-pro12%
llama-4-70b (shadow)5%