++++
UNIFIED INFERENCE GATEWAY
Inference Infrastructure
for Cost-Efficient AI
A unified inference gateway built for individual developers, startups, and growing teams — the same frontier model endpoints at one transparent rate, with nothing trained on your prompts.
No credit card required
PRICING · PER 1M TOKENS
Every model, a fraction of the price
FerrixOpenRouter list price
26 of 26 models
claude-fable-5
Anthropic
INPUT / 1M−80%
$2.00
$10.00
OUTPUT / 1M−80%
$10.00
$50.00
CACHE READ / 1M−80%
$0.20
$1.00
CACHE WRITE (5-MINUTE CACHE) / 1M−80%
$2.50
$12.50
CACHE WRITE (1-HOUR CACHE) / 1M−80%
$4.00
$20.00
VS OPENROUTER−80%
claude-haiku-4-5
Anthropic
INPUT / 1M−80%
$0.20
$1.00
OUTPUT / 1M−80%
$1.00
$5.00
CACHE READ / 1M−80%
$0.02
$0.10
CACHE WRITE (5-MINUTE CACHE) / 1M−80%
$0.25
$1.25
CACHE WRITE (1-HOUR CACHE) / 1M−80%
$0.40
$2.00
VS OPENROUTER−80%
claude-opus-4-6
Anthropic
INPUT / 1M−80%
$1.00
$5.00
OUTPUT / 1M−80%
$5.00
$25.00
CACHE READ / 1M−80%
$0.10
$0.50
CACHE WRITE (5-MINUTE CACHE) / 1M−80%
$1.25
$6.25
CACHE WRITE (1-HOUR CACHE) / 1M−80%
$2.00
$10.00
VS OPENROUTER−80%
claude-opus-4-7
Anthropic
INPUT / 1M−80%
$1.00
$5.00
OUTPUT / 1M−80%
$5.00
$25.00
CACHE READ / 1M−80%
$0.10
$0.50
CACHE WRITE (5-MINUTE CACHE) / 1M−80%
$1.25
$6.25
CACHE WRITE (1-HOUR CACHE) / 1M−80%
$2.00
$10.00
VS OPENROUTER−80%
claude-opus-4-8
Anthropic
INPUT / 1M−80%
$1.00
$5.00
OUTPUT / 1M−80%
$5.00
$25.00
CACHE READ / 1M−80%
$0.10
$0.50
CACHE WRITE (5-MINUTE CACHE) / 1M−80%
$1.25
$6.25
CACHE WRITE (1-HOUR CACHE) / 1M−80%
$2.00
$10.00
VS OPENROUTER−80%
claude-sonnet-4-6
Anthropic
INPUT / 1M−80%
$0.60
$3.00
OUTPUT / 1M−80%
$3.00
$15.00
CACHE READ / 1M−80%
$0.06
$0.30
CACHE WRITE (5-MINUTE CACHE) / 1M−80%
$0.75
$3.75
CACHE WRITE (1-HOUR CACHE) / 1M−80%
$1.20
$6.00
VS OPENROUTER−80%
claude-sonnet-5
Anthropic
INPUT / 1M−80%
$0.40
$2.00
OUTPUT / 1M−80%
$2.00
$10.00
CACHE READ / 1M−80%
$0.04
$0.20
CACHE WRITE (5-MINUTE CACHE) / 1M−80%
$0.50
$2.50
CACHE WRITE (1-HOUR CACHE) / 1M−80%
$0.80
$4.00
VS OPENROUTER−80%
deepseek-v3.2
DeepSeek
INPUT / 1M−64%
$0.0957
$0.269
OUTPUT / 1M−62%
$0.15
$0.40
Cache writes: not charged
VS OPENROUTER−62%
gemini-2.5-flash
Google
INPUT / 1M−80%
$0.06
$0.30
OUTPUT / 1M−80%
$0.50
$2.50
CACHE READ / 1M−80%
$0.006
$0.03
VS OPENROUTER−80%
gemini-2.5-pro
Google
INPUT / 1M−80%
$0.25
$1.25
OUTPUT / 1M−80%
$2.00
$10.00
CACHE READ / 1M−80%
$0.025
$0.125
VS OPENROUTER−80%
gemini-3.1-pro
Google
INPUT / 1M−80%
$0.40
$2.00
OUTPUT / 1M−80%
$2.40
$12.00
CACHE READ / 1M−80%
$0.04
$0.20
VS OPENROUTER−80%
gemini-3.5-flash
Google
INPUT / 1M−80%
$0.30
$1.50
OUTPUT / 1M−80%
$1.80
$9.00
CACHE READ / 1M−80%
$0.03
$0.15
VS OPENROUTER−80%
Page 1 of 3
Average saving across shown models: −77% · No minimums · No platform fee · Volume discounts above $2k/mo
WHAT PEOPLE SAY
Switched for the price. Stayed for everything else.
[ PRICE ]
"Same GPT-4o and Claude endpoints we were already calling, cut roughly in half. It paid for the migration in the first week."
DR
Dani Rowe
Staff Engineer
[ RELIABILITY ]
"We've never seen a blip — even during our biggest traffic spikes, latency stays flat and requests just go through. It's the most boring part of our stack, which is exactly what you want."
MK
Marcus Kang
Head of Platform
[ SWITCHING ]
"One base URL change and our SDK just worked. No rewrites, no new client — we were live in an afternoon."
PS
Priya Sharma
CTO
DROP-IN MIGRATION
Change one line.
Keep your stack.
Point your OpenAI-compatible client at Ferrix's base URL. Same request shape, same SDKs, same tooling — every model behind one endpoint.
WORKS WITH
openai · anthropic · langchain · llamaindex · vercel ai sdk
openai · anthropic · langchain · llamaindex · vercel ai sdk
CLIENT.PY
# before
client = OpenAI(
base_url="https://api.openai.com/v1",
api_key=KEY,
)
# after — that's the whole migration
client = OpenAI(
base_url="https://api.ferrix.sh/v1",
api_key=KEY,
)
$ curl api.ferrix.sh/v1/models
→ 142 models · 99.99% uptime
++++
SPEND CONTROLS
Spend controls that work before the invoice
[ CAP ]
Budget caps
Hard monthly ceilings per org, project, or key — requests stop before you overspend.
[ ALERT ]
Spend alerts
Real-time notifications at the thresholds you set, before the bill surprises you.
[ USAGE ]
Usage analytics
Per-model, per-key, per-project cost breakdowns, exportable anytime.
[ PRIVATE ]
No data training
We never train on, or retain, your prompts or completions. Ever.
RELIABILITY
99.99%
UPTIME, SLA-BACKED
180ms
P50 TIME-TO-FIRST-TOKEN
SOC 2
TYPE II · ZERO PROMPT RETENTION
Change one line tonight. Watch the bill drop.
Start with $5 in free credit — no card, no minimums.
++