01 / 7
OpenAI
GPT-5.5 Pro
General reasoning
MMLU92.4
SWE-bench71.2
GPQA67.8
1M ctx
$15.00$7.50/1M
−50%
GPT-5.5
CLAUDE
GEMINI
LLAMA
GROK
DEEPSEEK
MISTRAL
Anvat · 18 models live
Everymodel.Onekey.Halftheprice.
A drop-in OpenAI & Anthropic compatible proxy. Stream from GPT-5.5, Claude Opus 4.7, Gemini 3.5, Llama 4, Grok, DeepSeek and Mistral — through one endpoint.
Start with $2 free credit →
● LIVE$0.120/ 1M input−47%
Scroll ↓
01 / Pricing Engine
Engineered for efficiency.
We aggregate demand at industrial scale to secure bulk-tier rates from every provider. We pass 100% of that discount directly to your API calls. No markup. No middleman.
Monthly volume200M tokens
You save / month
$1477.50
OpenAI
GPT-5.5
−50%
Official API$0.00
Anvat$0.00
Anthropic
Claude Opus 4.7
−50%
Official API$0.00
Anvat$0.00
Google
Gemini 3.5 Pro
−58%
Official API$0.00
Anvat$0.00
xAI
Grok-4
−58%
Official API$0.00
Anvat$0.00
03 / Mechanism
Three steps. Zero friction.
STEP 01
Sign in once.
Continue with Google. Get $2 in free credit instantly — no card, no subscription, no expiry.
STEP 02
One unified API key.
Drop-in OpenAI and Anthropic SDK compatibility. Switch between GPT-5.5, Claude, Gemini by changing one string.
STEP 03
Pay only what you use.
Per-token billing across every provider. Real-time usage in your dashboard. Top up any time, no monthly minimum.
04 / Unified API
One endpoint.
Every frontier.
Drop-in compatible with the OpenAI and Anthropic SDKs. Switch from GPT-5 to Claude to Gemini to Llama by changing a single string. Streaming, function calling, vision — supported across every model.
- OpenAI-compatible JSON across every provider
- Anthropic /v1/messages support out-of-the-box
- Native streaming · function calling · vision
1import OpenAI from "openai";
2
3const ai = new OpenAI({
4 apiKey: process.env.PROXY_KEY,
5 baseURL: "http://api.anvat.app/v1",
6});
7
8// Same SDK. Different provider — change one string.
9const response = await ai.chat.completions.create({
10 model: "claude-opus-4-8",
11 messages: [{ role: "user", content: "Refactor this kernel." }],
12});
13
14console.log(response.choices[0].message.content);
Streaming · 142ms TTFT
05 / Pricing
Honest plans. No lock-in.
Starter
$10/mo
5M tokens included
For local development and prototypes.
- →Access to all base models
- →OpenAI- and Anthropic-compatible SDKs
- →Community Discord support
Recommended
Builder
$49/mo
60M tokens included
Production-grade infrastructure for shipping.
- →All frontier models · GPT-5.5, Opus 4.7
- →Priority routing queues
- →99.95% uptime SLA
- →Streaming · function calling
Scale
$299/mo
500M tokens included
High-throughput inference for teams.
- →Dedicated endpoints
- →24/7 engineer-led support
- →Custom fine-tuning
- →Volume discounts beyond cap
06 / Transparency
We're cheaper because the math says so.
01
Wholesale frontier-model access
We aggregate demand at industrial scale to secure bulk-tier rates from every provider — frontier weights, no per-request markup.
02
Optimised routing
Every request hits the cheapest qualified endpoint for the requested model — same weights, lower latency, lower cost.
03
Zero middleman markup
Pay the raw token cost plus a thin operations margin. Spend is itemised down to the request.
Routing diagram
07 / Builders
Shipping on Anvat.
Use case · 01
AI coding agents
Hover to reveal →
Route Claude 4.7 for planning, GPT-5.5 for refactors, DeepSeek for cheap completions — one key.
Opus 4.7GPT-5.5DeepSeek
— Mira, founder, ShipAgent
Use case · 02
RAG pipelines
Hover to reveal →
Embeddings on Gemini, reranking on Llama, synthesis on Opus. One key. One bill.
GeminiLlama 4Opus 4.7
— Yuki, ML eng, ContextLab
Use case · 03
Autonomous agents
Hover to reveal →
Plan with Opus, act with Grok, reflect with GPT — burning 200M tokens/day at half the cost.
Opus 4.7Grok-4GPT-5.5
— Renan, agent lead, OrionOS
Use case · 04
Content generation
Hover to reveal →
Long-form drafts on Gemini 2M context, polish on Opus. Same quality at 50% spend.
GeminiOpus 4.7
— Joon, head of content, Slate
Use case · 05
Customer support bots
Hover to reveal →
Tier-1 on Llama for $0.50/1M, escalation to GPT-5.5 — transparent fallbacks.
Llama 4GPT-5.5
— Priya, CTO, Lighthouse
Use case · 06
Voice & multimodal
Hover to reveal →
Vision on Gemini 3.5, ASR via integrated providers, all in one streaming endpoint.
GeminiGPT-5.5
— Tomas, audio eng, Whisp
09 / Operators
Engineers who ship.
“Cut our inference bill by 54% in one afternoon. Drop-in.”
Lena · CTO, Threadlight
“Switching models is a string change. That's the whole pitch.”
Devin · founder, Trace AI
“Finally a proxy that doesn't tax me to switch providers.”
Karan · staff eng, Northway
“Routed 4B tokens last month. Bill was lower than a solo OpenAI run.”
Hua · ML lead, Codex.io
“Cut our inference bill by 54% in one afternoon. Drop-in.”
Lena · CTO, Threadlight
“Switching models is a string change. That's the whole pitch.”
Devin · founder, Trace AI
“Finally a proxy that doesn't tax me to switch providers.”
Karan · staff eng, Northway
“Routed 4B tokens last month. Bill was lower than a solo OpenAI run.”
Hua · ML lead, Codex.io
“Latency is identical. Cost is half. I don't know how, I don't care.”
Sofia · principal, Helio
“The dashboard's transparency is the most honest thing in AI infra.”
Mateus · indie · Otto
“Opus, GPT, Gemini — one SDK. No more wrapper hell.”
Aiko · platform, Mesh
“Default choice for every new prototype now.”
Owen · founder, Quill
“Latency is identical. Cost is half. I don't know how, I don't care.”
Sofia · principal, Helio
“The dashboard's transparency is the most honest thing in AI infra.”
Mateus · indie · Otto
“Opus, GPT, Gemini — one SDK. No more wrapper hell.”
Aiko · platform, Mesh
“Default choice for every new prototype now.”
Owen · founder, Quill
10 / FAQ
Common questions.
We aggregate frontier-model capacity at scale and pass that wholesale rate through to you with a thin operations margin — no per-request markup, no hidden multiplier.
Everymodel.Onekey.Halftheprice.
Sign in with Google in under 30 seconds. $2 free credit on the house — no card required.
Claim your free credit →