INTELLIGENT ROUTING

One endpoint. The best execution.

Set the constraints that matter. Liquid finds the right path for every request.

Constraint-aware routing

4 eligible models / 17

OPTIMAL EXECUTION

◈

Qwen CoderAlibaba · US East

Google Vertex · US East

$1.58/ 1M output tokens

126mstime to first token

Model / provider	Quality	Output / 1M	TTFT	Eligibility
◈ Qwen CoderAlibaba · US East via Google Vertex	90	$1.58	126 ms	Eligible
✦ Gemini ProGoogle · US West via Cerebras	93	$9.84	146 ms	Eligible
◎ GPT-5.6 SolOpenAI · US East via Together	96	$10.48	164 ms	Eligible
✳ Claude SonnetAnthropic · US West via Fireworks	94	$13.92	182 ms	Eligible
∞ Llama 4Meta · US West via AWS Bedrock	82	$0.78	94 ms	Below quality target
N Nova ProAmazon · US Central via Azure AI	83	$2.49	134 ms	Below quality target
≈ DeepSeek V4DeepSeek · US East via Groq	89	$1.62	218 ms	Below quality target
✳ Claude OpusAnthropic · US East via Azure AI	98	$23.76	242 ms	Over price ceiling
M MinistralMistral · Nordics via Fireworks	76	$0.29	64 ms	Below quality target · Outside data residency
✦ Gemini FlashGoogle · Germany via Nebius	84	$1.24	86 ms	Below quality target · Outside data residency
◎ GPT MiniOpenAI · UK via Groq	85	$1.72	112 ms	Below quality target · Outside data residency
✳ Claude HaikuAnthropic · Canada via AWS Bedrock	85	$3.68	104 ms	Below quality target · Outside data residency
M Mistral LargeMistral · Germany via Cerebras	86	$3.81	142 ms	Below quality target · Outside data residency
◈ Qwen MaxAlibaba · Singapore via Google Vertex	88	$4.12	156 ms	Below quality target · Outside data residency
K Kimi K2Moonshot · Japan via Fireworks	88	$2.32	196 ms	Below quality target · Outside data residency
Z GLM 5Zhipu · Singapore via Together	87	$1.98	208 ms	Below quality target · Outside data residency
C Command R+Cohere · Canada via Nebius	81	$4.62	188 ms	Below quality target · Outside data residency

Your routing configuration

const response = await liquid.responses.create({
  "model": "liquid/auto",
  "objective": {
    "quality": 0.9,
    "max_price_per_million": 15,
    "max_ttft": 250,
    "min_context": 128000,
    "optimize": "balanced"
  },
  "constraints": {
    "jurisdiction": "US",
    "residency": true,
    "excluded_providers": []
  }
});

SMART ORDER ROUTING

Best execution for every token.

Your application requests / responses

Liquid Router PRICE × LATENCY × QUALITY

Fireworks$2.21 · 180msBEST EXECUTION

Together$1.98 · 242ms

Cerebras$2.44 · 91ms

Groq$2.36 · 114ms

Best price within the 200 ms latency target

LIQUID AGENT RUNTIME

Built for agents
that think a lot.

Reasoning loops, parallel rollouts, and long-running tasks. Keep inference and environment execution close.

01Reason

02Search

03Call tool

04Evaluate

05Simulate

06Verify

07Answer

Inference 81%Environment 14%Network 5%

Seven steps. One execution layer.

Routing tests use illustrative benchmarks and provider quotes. This frontend does not send requests to model providers.

One endpoint. The best execution.

4 eligible models / 17

Best execution for every token.

Built for agentsthat think a lot.

Built for agents
that think a lot.