Choose the Right Model for Your Use Case

TokenHub gives you access to dozens of models across OpenAI, Anthropic, Google, Mistral, and other providers through a single API. You can specify a model by its full provider-prefixed name, use a TokenHub alias that abstracts the provider away, or let TokenHub choose automatically based on your routing strategy.

How model selection works

There are three ways to specify a model in your request:

Provider-prefixed name — explicitly targets a specific model from a specific provider: openai/gpt-4o, anthropic/claude-3-5-sonnet, google/gemini-1.5-pro
TokenHub alias — a stable identifier like premium, balanced, or economy that maps to the best available model in that tier at request time
Auto-routing — omit the model field entirely and set X-Inferoute-Strategy: cost or X-Inferoute-Strategy: latency to let TokenHub decide

Model categories

Premium

Highest quality for complex reasoning, long-form writing, and multi-step tasks. Best when output quality directly affects user experience.

openai/gpt-4o
anthropic/claude-3-5-sonnet
Alias: premium

Balanced

Strong capability at moderate cost. A good default for most production workloads where you want solid quality without paying for top-tier.

openai/gpt-4o-mini
anthropic/claude-3-haiku
Alias: balanced

Economy

Fastest and cheapest. Ideal for classification, extraction, summarization, and any task where throughput matters more than nuance.

openai/gpt-4o-mini
anthropic/claude-3-haiku
google/gemini-flash
Alias: economy

Long context

For documents, codebases, and conversation histories that exceed standard context windows.

anthropic/claude-3-5-sonnet (200k tokens)
google/gemini-1.5-pro (1M tokens)
Alias: long-context

Recommended models by use case

Use case	Recommended model	Why
Complex reasoning, analysis	`openai/gpt-4o`, `anthropic/claude-3-5-sonnet`	Highest instruction-following and reasoning accuracy
Chatbots, customer support	`openai/gpt-4o-mini`, `anthropic/claude-3-haiku`	Low latency, cost-effective at high volume
Document summarization	`anthropic/claude-3-5-sonnet`, `google/gemini-1.5-pro`	Long context windows handle full documents
Classification / labeling	`openai/gpt-4o-mini`, `google/gemini-flash`	Simple tasks don’t need expensive models
Embeddings (search, RAG)	`openai/text-embedding-3-small`, `openai/text-embedding-3-large`	Purpose-built for vector representations
Code generation	`openai/gpt-4o`, `anthropic/claude-3-5-sonnet`	Strong code comprehension and generation

Model tier comparison

Premium
Balanced
Economy

Premium models deliver the best output quality across reasoning, writing, and instruction-following. Use them when accuracy and response quality are critical.Models: openai/gpt-4o, anthropic/claude-3-5-sonnetAlias: premium

python

response = client.chat.completions.create(
    model="openai/gpt-4o",
    messages=[
        {"role": "user", "content": "Analyze the trade-offs of microservices vs monolithic architecture for a startup with 5 engineers."},
    ],
)

node.js

const response = await client.chat.completions.create({
  model: "openai/gpt-4o",
  messages: [
    {
      role: "user",
      content:
        "Analyze the trade-offs of microservices vs monolithic architecture for a startup with 5 engineers.",
    },
  ],
});

Typical cost: ~

5–

15 per 1M input tokens

Balanced models cover most production workloads well. They handle nuanced tasks at a fraction of premium model cost.Models: openai/gpt-4o-mini, anthropic/claude-3-haikuAlias: balanced

python

response = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[
        {"role": "user", "content": "Draft a professional out-of-office email reply."},
    ],
)

node.js

const response = await client.chat.completions.create({
  model: "openai/gpt-4o-mini",
  messages: [
    {
      role: "user",
      content: "Draft a professional out-of-office email reply.",
    },
  ],
});

Typical cost: ~

0.15–

0.40 per 1M input tokens

Economy models are optimized for throughput and cost. Use them for tasks with clear, bounded outputs where you process high volumes.Models: openai/gpt-4o-mini, anthropic/claude-3-haiku, google/gemini-flashAlias: economy

python

response = client.chat.completions.create(
    model="economy",  # TokenHub alias — routes to cheapest available
    messages=[
        {"role": "user", "content": "Classify this support ticket as: billing, technical, or general."},
    ],
)

node.js

const response = await client.chat.completions.create({
  model: "economy", // TokenHub alias — routes to cheapest available
  messages: [
    {
      role: "user",
      content:
        "Classify this support ticket as: billing, technical, or general.",
    },
  ],
});

Typical cost: ~

0.07–

0.25 per 1M input tokens

Provider-agnostic code with aliases

When you hardcode a provider-prefixed model name like openai/gpt-4o, swapping providers requires a code change. TokenHub aliases let you decouple your application from any specific provider.

python

# Hardcoded — requires code change to switch providers
model = "openai/gpt-4o"

# Alias — TokenHub resolves to the best model in this tier
# You can update the alias mapping in the dashboard without touching code
model = "premium"

You can configure what each alias resolves to in the Model Aliases section of your dashboard. This lets you A/B test new models, react to provider outages, or optimize costs — all without deploying code changes.

Embeddings

For embeddings, specify the model directly. TokenHub passes the request to the provider without modification.

python

response = client.embeddings.create(
    model="openai/text-embedding-3-small",
    input="TokenHub routes your LLM requests intelligently.",
)

embedding = response.data[0].embedding
print(f"Embedding dimensions: {len(embedding)}")  # 1536 for text-embedding-3-small

Model	Dimensions	Best for
`openai/text-embedding-3-small`	1536	High-volume search, RAG pipelines where cost matters
`openai/text-embedding-3-large`	3072	Highest accuracy retrieval, semantic similarity

Documentation Index

​How model selection works

​Model categories

Premium

Balanced

Economy

Long context

​Recommended models by use case

​Model tier comparison

​Provider-agnostic code with aliases

​Embeddings

How model selection works

Model categories

Recommended models by use case

Model tier comparison

Provider-agnostic code with aliases

Embeddings