01 / Unified
One API. Every model.
Swap between GPT, Claude, Gemini, DeepSeek, Llama and 50+ models without rewriting a single line. OpenAI-compatible by default.
OneAPIforeverymodelyourstackwillneed.
›
The unified inference layer for GPT, Claude, Gemini, DeepSeek, Llama and 50+ models. Transparent pricing, predictable latency, one key.
All systems operational · 99.99% uptime
OpenAI
GPT-4o
Input · per 1M tokens
Official$5.00
TokenLayer$3.50-30%
Output · per 1M tokens
Official$15.00
TokenLayer$10.50-30%
Average savings: 30% across all models
8 models · Auto-rotating · Use arrows or dots to navigate
01
Edge uptime
0.00%
02
Models in catalog
0+
Capabilities
Theinfrastructurelayerbetweenyouandeverymodel.
Everything you need to ship reliable AI products — without managing six provider contracts.
How it works
Three steps from signup to scale.
est. setup time
≈ 4 minutes
01
Create your account.
Sign up in 60 seconds. No credit card. $5 of free credits to play with every model.
› signup —› tokenlayer.net/get-started
02
Grab your API key.
Generate a key, set a monthly cap, and we'll handle quota across every provider.
› TOKENLAYER_KEY=tl_live_a1b2c3d4...
03
Start building.
Drop-in compatible with the OpenAI SDK. Change baseURL — keep your existing code.
› client.chat.completions.create({ model: "auto" })
The model catalog
Every frontier model. Always up to date.
From the day a new model ships, it's available on TokenLayer — with the same SDK call, same billing, same observability.
The model catalog
Every frontier model. Always up to date.
From the day a new model ships, it's available on TokenLayer — with the same SDK call, same billing, same observability.
14 added this month
OpenAI
GPT
Flagship reasoning and multimodal models.
4o4o-minio1o3
model.gpt
Anthropic
Claude
Long-context and coding-grade quality.
3.5 Sonnet3.5 Haiku3 Opus
model.claude
Google
Gemini
Native multimodal with 2M+ token context.
2.0 Pro2.0 Flash1.5 Pro
model.gemini
DeepSeek
DeepSeek
Frontier-grade reasoning at a fraction of the cost.
V3R1Coder
model.deepseek
Meta
Llama
Open-weight models optimized for speed.
3.3 70B3.2 90BGuard
model.llama
+ 50 more
And more
Every notable provider, one endpoint.
MistralQwenxAICohere
model.andmore
+ More daily
Every model. The day it ships.
We add new frontier models within hours of release, with full parity for streaming, tools and structured outputs.
SCROLL
7 MODELS
Calculate your savings
See how much you'll save.
Adjust the sliders to estimate your monthly costs based on real provider pricing.
Tip: Drag the sliders or click any value to type your own
Select Model
1001.0M
10032,000
Total tokens / mo
20.0M
OpenAI
$5/$15 / 1M
Your monthly cost
Direct from provider
$180
With TokenLayer-80%
$36.00
You save / month
$144
You save / year
$1.73k
Estimates based on 60/40 input/output split. Prices from pricepertoken.com
Developer experience
Fiveminutestoyourfirsttoken.
OpenAI-compatible API, first-class SDKs, native streaming, structured outputs, tools, and observability. The boring parts, done right.
- Drop-in replacement for OpenAI SDK
- Streaming + structured outputs out of the box
- Automatic retries, fallback chains, request signing
- Per-request tracing with full payloads
app.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
// One client, any model. import { TokenLayer } from "@tokenlayer/sdk"; const tokenlayer = new TokenLayer({ apiKey: process.env.TOKENLAYER_KEY, }); const res = await tokenlayer.chat.completions.create({ model: "auto", // routes to cheapest model messages: [{ role: "user", content: "Summarize Q3." }], fallback: ["claude-3.5-sonnet", "gpt-4o"], stream: true, }); for await (const chunk of res) { process.stdout.write(chunk.choices[0]?.delta?.content ?? ""); }
● 200 OK→ claude-3.5-sonnet$0.0021
642ms · 1,284 tokPricing
Transparent.Nosurprises.
Pay only for what your models consume — plus a flat per-seat fee. No markups on tokens, ever.
Token costs passed through at provider list price. See full pricing →
What you can do
Buildmore.Spendless.Sleepbetter.
Concrete outcomes you can expect after pointing your stack at TokenLayer.
Avg. 41% cost reduction in week one
Questions
Common questions, honest answers.
Still curious? Our team responds to every email within a business day.
Yes. Point your base URL at api.tokenlayer.net/v1 and use any model name. Our endpoints are 100% compatible with the OpenAI Chat Completions and Responses APIs — including streaming, tool calling, and structured outputs.
Ready when you are
Buildwitheverymodelbeforelunch.
$5 of credits when you sign up. No credit card. Cancel any time.
SOC 2 Type IIGDPR · HIPAA · BAA99.99% Uptime SLA14 global PoPs