uptime 99.994%/28 models · 10 providers
One key.
Every model.
Zero rate ceilings.
Forge is an OpenAI-compatible routing gateway. Point your SDK at us and stream from GPT-5.6, Claude Opus, Grok, DeepSeek, Kimi and 20+ more — with automatic key pooling and millisecond failover.
22.1B
tokens routed
573ms
avg first-token
99.99%
routing SLA
gateway.forge/edge · iad1
live08:36:24200POST /v1/chat/completionsclaude-sonnet-4-5iad1214ms
08:36:22200POST /v1/chat/completionsdeepseek-r1sfo1891ms
08:36:21200POST /v1/chat/completionsgpt-5.6-lunafra1342ms
08:36:20200GET /v1/models-iad112ms
08:36:18200POST /v1/chat/completionsgrok-4.5iad1512ms
08:36:17429POST /v1/chat/completionskimi-k3hkg11024ms
08:36:16200POST /v1/chat/completionskimi-k3hkg163ms
08:36:14200POST /v1/chat/completionsgemini-3-proiad1421ms
08:36:13200POST /v1/embeddingstext-embed-3iad187ms
08:36:12200POST /v1/chat/completionsclaude-sonnet-4-5iad1214ms
08:36:11200POST /v1/chat/completionsdeepseek-r1sfo1891ms
08:36:09200POST /v1/chat/completionsgpt-5.6-lunafra1342ms
08:36:08200GET /v1/models-iad112ms
08:36:07200POST /v1/chat/completionsgrok-4.5iad1512ms
08:36:05429POST /v1/chat/completionskimi-k3hkg11024ms
08:36:04200POST /v1/chat/completionskimi-k3hkg163ms
08:36:03200POST /v1/chat/completionsgemini-3-proiad1421ms
08:36:01200POST /v1/embeddingstext-embed-3iad187ms
Routing toopenaianthropicxaideepseekmoonshotgoogletencentxiaomiminimaxglm
/* 01 — integration */
Drop-in for the OpenAI SDK.
Change two lines. Keep your streaming, tool-calls, JSON mode. Every request is observable in the Forge dashboard and billed against a single balance.
- base_url
- https://forge-gateway-api.fly.dev/v1
- auth
- Bearer fg-*
- endpoints
- /chat/completions · /embeddings · /models
- region
- iad1 · sfo1 · fra1 · hkg1
agent.pycurl.shnode.ts
from openai import OpenAI
client = OpenAI(
api_key="fg-demotoken1234567890",
base_url="https://forge-gateway-api.fly.dev/v1",
)
response = client.chat.completions.create(
model="deepseek-r1",
messages=[{"role": "user", "content": "Explain load balancing."}],
stream=True,
)
for chunk in response:
print(chunk.choices[0].delta.content, end="")▸ streaming · 1.2s to first token · routed deepseek-r1 via iad1
/* 02 — catalog */
Frontier models, one endpoint.
OpenAIFREE
gpt-5.6-luna
1.05M ctx$1.00 / $6.00
OpenAIFREE
gpt-5.6-sol
1.05M ctx$5.00 / $30.00
OpenAIFREE
gpt-5.6-terra
1.05M ctx$2.50 / $15.00
OpenAIFREE
gpt-5.5
128K ctx$5.00 / $40.00
OpenAIFREE
gpt-5.3-codex
400K ctx$1.75 / $14.00
AnthropicFREE
claude-opus-4-5
200K ctx$5.00 / $25.00
AnthropicFREE
claude-sonnet-4-5
200K ctx$3.00 / $15.00
AnthropicFREE
claude-haiku-4-5
200K ctx$1.00 / $5.00
/* 03 — plans */