v0.1 · private beta open

One API.
Every model. Routed to the best.

The open OpenRouter alternative — a unified inference gateway and live leaderboard for LLMs and multimodal AI. OpenAI-compatible, 300+ models, and cost / latency / quality-aware routing — without the vendor lock-in.

$5 free credits · no card · works with the OpenAI SDK

312
models routed
4.7M
tokens / sec
28
providers
142ms
p50 latency
99.98%
uptime (30d)

Routing across the entire frontier — and every serious open model

OpenAIAnthropicGoogleMetaMistralDeepSeekQwenxAICohereGroqTogetherFireworksReplicatePerplexityOpenAIAnthropicGoogleMetaMistralDeepSeekQwenxAICohereGroqTogetherFireworksReplicatePerplexity
What ships in v1

A production gateway, not a wrapper.

Built for teams that ship AI products — with the primitives you'd otherwise build yourself.

01 / gateway

OpenAI-compatible, drop-in

Point your existing OpenAI SDK at api.llmcloud.ai and you're done. Streaming, tools, structured output, vision, PDFs — all pass through unchanged.

02 / routing

Commission-based smart routing

A prompt classifier picks the cheapest model that meets your quality bar. Cascade to a stronger one only when the first refuses. Save 40–80% on the same task.

03 / multimodal

First-class image, audio, video

FLUX, Nano Banana, Imagen, Veo, Whisper, ElevenLabs — one shape for text, images, speech and video. Not a text-only surface with attachments bolted on.

04 / leaderboard

Live, honest rankings

Public benchmarks + our own evals + real gateway traffic. Filter by task, price ceiling, latency budget, context window — not a single averaged Elo.

05 / observability

Traces, evals, guardrails

Every request is logged with cost, latency, and token usage. PII redaction, prompt-injection filters, and per-key spend limits ship in the box.

06 / control

BYOK, fallbacks, provider prefs

Bring your own provider keys and pay providers directly. Set fallback chains, exclude regions, cap per-request cost. You stay in control.

Live leaderboard

Rankings that actually match your workload.

Not one averaged Elo. Combine public benchmarks (MMLU-Pro, GPQA, LiveBench, SWE-Bench), our task-specific evals, and real traffic from the gateway.

#
Model
Quality
Tok/s
TTFT
$ / Mtok
01
claude-opus-4.5
Anthropic
94.2
82
320
$15.00
02
gpt-5.5
OpenAI
93.8
74
410
$12.50
03
gemini-3.1-pro
Google
92.1
118
180
$7.00
04
grok-5
xAI
90.6
95
260
$6.00·
05
deepseek-v4
DeepSeek
89.9
140
210
$0.55
06
qwen3-max
Alibaba
88.4
128
240
$0.90
07
llama-4-405b
Meta
87.6
165
150
$2.40·
08
mistral-large-3
Mistral
86.1
110
280
$3.00

Preview data · full leaderboard ships with launch

Smart routing

Cheapest model that meets your bar. Every time.

A lightweight classifier reads each prompt and picks from a candidate pool you control. Fallback if the primary refuses or times out. Cascade if a cheap model isn't confident. You define the tradeoff — we do the work.

  • Cost-first
    Cheapest model above quality floor Q ≥ 0.85 for the task.
  • Latency-first
    P50 TTFT under 200ms — routed to the closest healthy provider.
  • Quality-first
    Frontier model, with a cheap draft-and-verify to cut spend.
  • Custom
    Write your own routing policy in JSON or a few lines of TypeScript.
POST /v1/chat/completions
{
  "model": "auto:cost",
  "messages": [...],
  "quality_floor": 0.85
}
deepseek-v4cheap
$0.55/M
gpt-5.5balanced
$12.50/M
claude-opus-4.5frontier
$15.00/M
Drop-in

Change one line. Ship.

If it speaks OpenAI, it speaks llmcloud. Streaming, tools, vision, audio, structured outputs — supported unchanged.

api.llmcloud.ai/v1
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.llmcloud.ai/v1",
  apiKey: process.env.LLMCLOUD_API_KEY,
});

const res = await client.chat.completions.create({
  model: "auto:cost",           // let llmcloud pick
  messages: [{ role: "user", content: "Explain MoE in one line." }],
  // optional: constrain the pool
  extra_body: { quality_floor: 0.85, max_latency_ms: 800 },
});
Pricing

Provider price + a thin platform fee.

No markup on tokens. No surprise inflation. Buy credits, or bring your own provider keys and pay them directly.

Hobby
Free
$5 credit to start
  • 300+ models
  • 5 RPS
  • Community support
  • Public leaderboard
Most teams
Pro
Pay as you go
5% platform fee
  • Smart routing
  • BYOK supported
  • Traces & evals
  • Team API keys
Enterprise
Custom
SOC 2 · SSO · VPC
  • Private routing pools
  • Dedicated OSS hosting
  • On-prem gateway
  • Support SLA

Ship AI without picking sides.

Join the beta. Get $5 in credits. One line of code from your first request.