Portkey Model Catalog
Browse, compare, and select from 1,600+ AI models across providers with real-time cost, latency, and capability benchmarks

Comprehensive Model Coverage

1,600+ models indexed across every major provider – OpenAI, Anthropic, Google, Mistral, Meta, Cohere, AWS Bedrock, Azure, and hundreds more – all queryable from a single interface. The catalog covers both open and closed source models and is updated continuously as new models are released.

• 1,600+ models indexed across 200+ providers.
• OpenAI, Anthropic, Google, Mistral, Meta, Cohere, AWS Bedrock, Azure, and more.
• Open and closed source models included.
• Updated continuously as new models are released.

Side-by-Side Comparison

Compare any set of models on cost per input and output token, latency, context window size, supported modalities, and published benchmark scores in a single view. Eliminate the manual effort of collating this data from individual provider documentation pages.

• Cost per input and output token comparison.
• Context window size across all models.
• Latency benchmarks side by side.
• Published benchmark score comparison.

Real-Time Pricing Data

Up-to-date pricing data for every model and provider, updated continuously and with historical price tracking so teams can monitor how provider pricing evolves over time and make cost-optimisation decisions based on accurate current data.

• Real-time price data updated continuously.
• Input and output token pricing per model.
• Cross-provider cost comparison in one view.
• Historical price tracking for trend analysis.

Latency Benchmarks

Measured p50 and p99 latency data from Portkey's global gateway infrastructure, reflecting real-world performance rather than provider-quoted figures. Updated continuously as infrastructure and model performance changes.

• p50 and p99 latency measured from global gateway infrastructure.
• Real-world performance data rather than provider estimates.
• Per-provider and per-model latency comparison.
• Updated continuously as performance changes.

Capability Filters

Filter the catalog by modality (text, vision, audio, embeddings), context window length, fine-tuning support, function calling capability, and compliance certifications to quickly narrow to the right model for a specific use case without reviewing every entry manually.

• Filter by modality: text, vision, audio, embeddings.
• Context window length filter.
• Fine-tuning support and function calling flags.
• Compliance certification filter for regulated use cases.

Stop Guessing – Choose Models With Data

The AI model landscape changes every week. New models are released, prices drop, and latency characteristics shift. Portkey's Model Catalog gives teams current, accurate data on cost, latency, and capabilities so model selection decisions are based on evidence rather than reputation or recency bias.

• 1,600+ models across 200+ providers in one searchable view.
• Real-time pricing and latency benchmarks updated continuously.
• Side-by-side capability comparison across any model set.
• Historical pricing trends for long-term cost planning.

One-Click Gateway Integration

Every model in the catalog is directly accessible via the Portkey AI Gateway. Switching from one model to another is a configuration change at the gateway layer – no application code changes, no redeployment, no new SDK integrations required. Teams can respond to new model releases or pricing changes within minutes.

• All catalogued models accessible via the Portkey AI Gateway.
• Model switching via configuration – no code changes required.
• Config-driven model selection per virtual key or team.
• Respond to new releases and pricing changes immediately.

A/B Testing via the Catalog

Combine the Model Catalog with the Portkey AI Gateway's A/B testing capability to split production traffic between two models and compare quality, latency, and cost on real requests before fully committing to a switch. Promote the better-performing model with a single configuration change.

Avoid Vendor Lock-In

Because every catalogued model is accessible through the same gateway API, teams are never locked into a single provider. Migrating between providers – or running multiple providers simultaneously for different workloads – requires no changes to application code and no new credentials exposed to application teams.

• No vendor lock-in – switch providers via configuration.
• Run multiple providers simultaneously for different workloads.
• Raw provider credentials never exposed to application teams.
• Provider migration without application redeployment.

Portkey Model Catalog Specifications: