Foundation model APIs
Programmatic LLM access
Direct API access to frontier and open-weight large language models. The building blocks for AI features in your own products.
-
#01
Anthropic API
· Anthropic Recommended FeaturedAnthropic's API for Claude models. Strongest models for code, long-document reasoning, and careful writing; native MCP support for tool integration; clean prompt-caching pricing.
AIMenta — Our pick when output quality matters more than feature breadth. Pair with OpenAI for voice/image and you cover the full surface.
Usage-based · Sonnet 4.6 ~US$3/M input tokens · API · Since 2023 -
#02
OpenAI API
· OpenAI Recommended FeaturedThe frontier-model API that launched the category. Best-in-class developer experience, broadest tool ecosystem, and the most widely benchmarked model family.
AIMenta — Default API for production AI features unless cost or residency forces otherwise. The developer surface and reliability remain a step ahead of competitors.
Usage-based · GPT-5 ~US$2.50/M input tokens · API · Since 2020 -
#03
AWS Bedrock
· Amazon RecommendedAWS's managed gateway to multiple foundation models — Claude, Llama, Mistral, Amazon Titan/Nova, and others — with IAM, VPC, and data residency controls suited for regulated enterprises.
AIMenta — For AWS-committed enterprises with data governance needs, Bedrock is usually the right answer despite the model lag and pricing premium.
Usage-based · Per-model pricing · API · Since 2023 -
#04
Azure OpenAI Service
· Microsoft RecommendedOpenAI's frontier models served through Azure with enterprise data controls, regional deployments, and integration with Azure AD. The standard answer for Microsoft-aligned enterprises.
AIMenta — For most APAC banks and regulated enterprises this is the cleanest path to GPT in production. Plan capacity carefully — PTU shortages bite at scale.
Usage-based · Per-model pricing · API · Since 2023 -
#05
Google AI / Vertex AI
· Google RecommendedGemini API via Google AI Studio (developers) or Vertex AI (enterprise). Strongest pricing and largest context windows in the frontier tier; native long-form video understanding.
AIMenta — Essential to evaluate for any long-context or multimodal use case. For pure text reasoning, the gap with Claude and GPT-5 has narrowed but not closed.
Usage-based · Gemini 3 Pro ~US$1.25/M input · API · Free tier · Since 2023 -
#06
Snowflake Cortex
· Snowflake RecommendedSnowflake's AI suite — LLM functions over your warehouse data, Cortex Search, Cortex Analyst (NL-to-SQL), and Document AI. Data stays inside Snowflake.
AIMenta — For Snowflake customers, Cortex is usually the right answer for AI on warehouse data. Avoids exporting to external LLM APIs.
Usage-based · Bundled with Snowflake credits · API · Since 2024 -
#07
Cohere
· Cohere Decent fitEnterprise-focused LLM provider with strong RAG and embedding models. Notable for private deployment options and a focus on regulated-industry customers.
AIMenta — Strong choice when retrieval quality and private deployment matter more than raw frontier reasoning. Their rerank model is class-leading.
Usage-based · Command R+ ~US$2.50/M input · API · Free tier · Since 2019 -
#08
Mistral La Plateforme
· Mistral AI Decent fitMistral's API with EU data residency by default. Open-weight options (Mistral Small, Mixtral) and closed frontier (Mistral Large 3). Strong code model with Codestral.
AIMenta — A serious option for EU operations. For APAC, the lack of regional residency limits its enterprise appeal.
Usage-based · Mistral Large ~US$2/M input · API · Free tier · Since 2023 -
#09
DeepSeek
· DeepSeek WatchChinese open-weight reasoning model with frontier-class performance at a fraction of the price. Strong for cost-sensitive APAC deployments where Chinese data residency is acceptable.
AIMenta — A genuine engineering achievement and a serious option for cost-sensitive use cases — but enterprise data governance must be evaluated case by case.
Freemium · Free chat; API ~US$0.14/M input tokens · API · Free tier · Since 2023