LLM Providers

A Survey of Model Providers Supported by {ellmer}

llm

A quick survey of the major players in the LLM space.

Published

March 22, 2026

An analyst should understand the landscape of LLM providers to make informed decisions about which models. Pay attention on operational and practical aspects that affect your work. Here’s what matters most.

  1. Model Capabilities for Summarization. How much text (tokens) the model can process without chunking (context window). Longer windows (e.g., 128K+ tokens).
  2. Per-Token/Input-Output Rates: Compare input and output costs.
  3. Ease of Integration and Reliability. Stick with models integrated into {ellmer} for seamless use.
  4. Data Privacy and Security:. Does the provider store your text data? For sensitive data ensure compliance with GDPR, HIPAA, or similar. Avoid providers that use data for training without opt-out.
  5. Scalability and Flexibility. Multiple models (e.g., lightweight for drafts, advanced for final summaries) let you optimize for cost vs. quality.

Categorizing the Landscape

{ellmer} 0.4.0 supports 17 providers. That’s a great list to understand the LLM landscape. We can group them into three categories based on their primary role:

Model Developers

Companies that primarily build and train their own LLMs.

  • Anthropic (Claude)
  • DeepSeek
  • Google (Gemini)
  • Mistral
  • OpenAI (GPT)

Platform/Cloud Distributors

Companies that host, integrate, or resell models from developers, often with added infrastructure or enterprise features.

  • AWS Bedrock
  • Azure OpenAI
  • Cloudflare
  • Databricks
  • GitHub Model Marketplace
  • Groq
  • Hugging Face
  • OpenRouter
  • Snowflake Cortex

Tools/Utilities

Open-source or specialized tools for running or integrating models, without core development.

  • Ollama
  • perplexity.ai
  • VLLM

Think of this breakdown as: Developers for innovation, Distributors for accessibility, Tools for flexibility/privacy.

Survey of Major LLM Providers

Let’s briefly survey the major LLM providers.

Companies that primarily build and train their own LLMs.

Who They Are: Founded by ex-OpenAI researchers, Anthropic focuses on building safe, beneficial AI. They’re known for prioritizing AI safety and ethical development.

Why Pay Attention: Rapidly growing with major partnerships (Amazon, Google, Microsoft), they’re pushing boundaries in AI safety and reasoning capabilities. Their models are frequently updated and competitive in benchmarks.

Capabilities for Summarization: Claude models offer some of the longest context windows in the industry and excel at complex reasoning and multi-step analysis, making them strong for nuanced qualitative data summarization.

Pricing & Integration: Competitive per-token pricing with batch processing discounts. Strong API reliability and enterprise features.

Data Privacy: Enterprise-grade privacy with US-only inference options; no data training without opt-out.

Scalability & Ecosystem: Best-in-class for agentic workflows and complex tasks; available via major cloud platforms (AWS Bedrock, Google Vertex, Microsoft Foundry) for multi-cloud deployments.

Business Model & Positioning: Anthropic monetizes through API consumption (per-token pricing), consumer subscriptions, enterprise licensing, and partnerships with cloud providers. Strong competitive advantage in reasoning/extended thinking capabilities and enterprise safety focus; positioned as premium provider targeting regulated industries and complex reasoning tasks.

Who They Are: Chinese AI startup focused on efficient, open-source models.

Why Pay Attention: Offers exceptional value with strong reasoning at low cost; growing rapidly.

Capabilities for Summarization: Models with good context windows and reasoning, competitive with premium providers.

Pricing: Significantly cheaper than OpenAI (~90% less).

Data Privacy: China-based; may not suit US/EU compliance needs.

Scalability & Ecosystem: Open-source focus with aggregator availability; good for budget teams.

Business Model & Positioning: DeepSeek (China-based) monetizes through freemium models with API consumption (~$0.27/$0.40 per million tokens for V3.2—80-90% cheaper than OpenAI). Market positioning: exceptional cost-to-capability ratio targeting budget-conscious teams, strong reasoning via distilled R1 models; disadvantage for US/EU enterprises requiring data sovereignty and compliance guarantees.

Who They Are: Google’s AI division, leveraging DeepMind research, offering models through Vertex AI platform.

Why Pay Attention: Backed by Google’s resources and research, they provide multimodal and multilingual capabilities with strong cloud integrations.

Capabilities for Summarization: Gemini models support multimodal input (text + images) and excel in multilingual tasks, useful for diverse data summarization.

Pricing: Competitive per-token rates.

Data Privacy: Strong Google Cloud compliance (SOC 2, HIPAA-eligible).

Scalability & Ecosystem: Deep GCP integrations (BigQuery, etc.); great for Google-centric workflows.

Business Model & Positioning: Google monetizes through per-API token consumption ($1.25-$15/M input tokens) and Vertex AI enterprise services with commitments; leverages Google Cloud infrastructure for customer lock-in. Competitive strengths: multimodal capabilities, 100+ language support, integration with BigQuery/workbenches, and backing by DeepMind research for continuous capability upgrades.

Who They Are: European AI company focused on open-source models.

Why Pay Attention: Strong on European languages and privacy compliance.

Capabilities for Summarization: Good context and multilingual support.

Pricing: Competitive rates.

Data Privacy: EU-based with GDPR focus.

Scalability & Ecosystem: Open-source options; enterprise deployments.

Business Model & Positioning: Mistral (EU-based) monetizes through API consumption (~$0.30-$2/M tokens), enterprise licensing, and custom model development services (Forge, advanced R&D). Market positioning: EU data residency compliance with GDPR enforcement, open-source model availability for private deployment, and strong European enterprise focus; advantage for regulations-sensitive organizations.

Who They Are: Founded by Elon Musk and a team of AI researchers, OpenAI is a leading independent AI company focused on advancing AGI (artificial general intelligence) safely.

Why Pay Attention: As pioneers in generative AI, they have massive influence, frequent model updates, and widespread adoption across industries. Their models set industry standards.

Capabilities for Summarization: GPT family models excel at complex reasoning and multi-step tasks, making them reliable for nuanced qualitative data summarization.

Pricing: Premium per-token rates, reflecting their cutting-edge performance.

Data Privacy: Zero data retention by default; strong compliance options (SOC 2, HIPAA).

Scalability & Ecosystem: Mature ecosystem with extensive documentation and integrations; ideal for production deployments despite higher costs.

Business Model & Positioning: OpenAI monetizes through: (1) ChatGPT subscriptions ($20+/month consumer), (2) API token consumption ($5-$25/M tokens depending on model tier), (3) enterprise licensing/agreements (Department of War contract, Amazon partnership). Market positioning: frontier model performance (GPT-5 series), industry-standard API, and brand reputation; disadvantage: premium pricing and vendor lock-in risk.

Companies that host, integrate, or resell models from developers, often with added infrastructure or enterprise features.

Who They Are: Amazon’s managed service for accessing various foundation models.

Why Pay Attention: Offers a wide range of models with strong enterprise features and AWS integrations.

Capabilities for Summarization: Access to top models like Claude and Llama; good for benchmarking and optimization.

Pricing: Pay-as-you-go or provisioned throughput.

Data Privacy: Strong compliance (HIPAA, SOC 2); no training on customer data.

Scalability & Ecosystem: Deep AWS integrations; excellent for large-scale, multi-model deployments.

Business Model & Positioning: Amazon’s multi-model platform monetizes through pay-per-use token consumption and provisioned throughput commitments; generates revenue while strengthening AWS ecosystem lock-in. Competitive strength lies in model diversity (100+ models), deep AWS integration, compliance certifications (HIPAA, FedRAMP), and cost optimization for enterprise customers at scale.

Who They Are: Microsoft’s cloud service offering exclusive access to OpenAI’s models within the Azure ecosystem.

Why Pay Attention: Combines OpenAI’s cutting-edge models with Microsoft’s enterprise infrastructure, compliance, and integrations—ideal for large organizations.

Capabilities for Summarization: Access to GPT family models with strong analytical performance, suitable for financial and qualitative data summarization.

Pricing: Provisioned throughput tied to Azure consumption (enterprise-focused).

Data Privacy: Top-tier compliance (HIPAA, FedRAMP) with regional data residency and encryption.

Scalability & Ecosystem: Seamless Microsoft 365/Power BI integrations and governance tools; best for Microsoft-centric enterprises.

Business Model & Positioning: Microsoft’s exclusive Azure distribution of OpenAI’s models generates revenue through provisioned/standard throughput pricing tied to Azure account consumption. Key differentiators: Microsoft 365 integration, priority access to latest models, regional data residency compliance, and bundling with Azure AI services for enterprises using Microsoft tech stacks.

Who They Are: Cloud infrastructure company expanding into AI with edge-deployed models.

Why Pay Attention: Focuses on low-latency, global inference for fast, accessible AI.

Capabilities for Summarization: Supports various open models with good context; cost-effective for moderate workloads.

Pricing: Flexible, often included in plans.

Data Privacy: Edge deployment minimizes data transfer.

Scalability & Ecosystem: Global POPs for low latency; integrates with Cloudflare tools.

Business Model & Positioning: Cloudflare monetizes LLM inference through edge-hosted models across 200+ POPs; included in Workers Pro plans or priced per-token for consumption. Unique positioning: ultra-low latency inference at the edge, developer-friendly integration, and data residency guarantees; competitive advantage for real-time, latency-sensitive applications.

Who They Are: Data platform company offering AI integrations in their lakehouse.

Why Pay Attention: Seamlessly combines data processing with AI for analysts.

Capabilities for Summarization: Native SQL/Python functions for summarization without data movement.

Pricing: Consumption-based on platform usage.

Data Privacy: Data stays in your lakehouse/VPC.

Scalability & Ecosystem: Integrates with Spark/Delta; ideal for Databricks users.

Business Model & Positioning: Databricks generates revenue through consumption-based pricing on its unified lakehouse platform (DBU/DSU units) for data storage, compute, and LLM serving integrated with data pipelines. Competitive advantage: data stays in lakehouse with native governance, seamless Spark/Delta Lake integration, and end-to-end lineage for analysts already invested in Databricks ecosystem.

Who They Are: GitHub’s emerging AI model hosting service.

Why Pay Attention: Convenient for developers already on GitHub.

Capabilities for Summarization: Access to models like Llama and Mistral.

Pricing: Mixed, tied to GitHub Enterprise.

Data Privacy: Standard GitHub policies.

Scalability & Ecosystem: Integrates with Copilot; early-stage.

Business Model & Positioning: GitHub monetizes model serving through GitHub Enterprise integrations and potential future revenue-sharing with model creators; currently early-stage with limited API infrastructure. Positioning: convenient for GitHub-native developer workflows with Copilot integration; disadvantages include unclear long-term SLA commitments and limited model diversity compared to competitors.

Who They Are: Hardware-focused startup with custom AI chips for fast inference.

Why Pay Attention: Pioneers in low-latency AI, offering speed advantages for real-time tasks.

Capabilities for Summarization: Supports fast models like Llama; ideal for quick, iterative summarization.

Pricing: Low-cost token rates.

Data Privacy: No training on user data.

Scalability & Ecosystem: Best for latency-critical workflows; limited model variety.

Business Model & Positioning: Groq monetizes through GroqCloud consumption (per-token pricing) leveraging its proprietary LPU (Language Processing Unit) hardware; recently raised $750M for scaling. Unique differentiator: 10x faster inference throughput via custom silicon; competitive advantage for latency-critical applications (real-time analytics, streaming); disadvantage: model diversity limited vs. GPU-based competitors.

Who They Are: Open-source AI community platform with model hosting.

Why Pay Attention: Massive ecosystem for experimenting with diverse models.

Capabilities for Summarization: 2M+ models, including specialists; good for research.

Pricing: Free tier + metered for compute.

Data Privacy: Depends on provider; open-source options for local use.

Scalability & Ecosystem: Community-driven; flexible for prototyping.

Business Model & Positioning: Hugging Face monetizes through: (1) Inference Endpoints compute ($0.60+/hour), (2) Inference API token consumption via partner providers ($0.02-$1.48/M tokens), (3) Enterprise Team/HF Compute plans ($20+/user/month). Strengths: massive open-source ecosystem (2M+ models), research community engagement, flexibility to run models locally; positioning as community-first, with enterprise services layered on top.

Who They Are: Aggregator providing unified access to multiple AI providers.

Why Pay Attention: Simplifies experimenting across providers without switching APIs.

Capabilities for Summarization: Routes to top models for cost/latency optimization.

Pricing: Variable, based on underlying providers.

Data Privacy: Inherits from chosen provider.

Scalability & Ecosystem: Fallback routing for reliability; no lock-in.

Business Model & Positioning: OpenRouter monetizes through margin on aggregated provider pricing (pay-as-you-go credits, no subscriptions); acts as a marketplace arbitrageur distributing requests across 60+ providers/300+ models. Strengths: price comparison tools, multi-provider uptime optimization, no vendor lock-in; positioning as neutral aggregator for cost optimization and reliability-conscious teams.

Who They Are: Snowflake’s AI features integrated into their data warehouse.

Why Pay Attention: Keeps data in-warehouse for seamless AI workflows.

Capabilities for Summarization: SQL functions for direct summarization.

Pricing: Tied to Snowflake consumption.

Data Privacy: Data stays secure in warehouse.

Scalability & Ecosystem: Ideal for Snowflake users; no data movement.

Business Model & Positioning: Snowflake monetizes through consumption-based pricing on Cortex AI functions within Snowflake warehouse credits, alongside data storage/compute. Competitive advantage: data stays in secure warehouse perimeter with native governance, seamless SQL integration (SUMMARIZE() functions), and unified structured/unstructured data analysis—optimal for existing Snowflake customers.

Open-source or specialized tools for running or integrating models, without core development.

Who They Are: Open-source tool for local AI model inference.

Why Pay Attention: Enables privacy-focused, offline AI without cloud costs.

Capabilities for Summarization: Runs various open models locally.

Pricing: Free and open-source.

Data Privacy: Complete local control; no cloud exposure.

Scalability & Ecosystem: Scales to your hardware; community integrations.

Business Model & Positioning: Ollama monetizes through free open-source software with potential cloud hardware upsells (faster inference via cloud account); future monetization via premium features and managed services announced. Positioning: local-first inference for data privacy, 40K+ integrations with dev tools; advantage for teams avoiding cloud dependencies; disadvantage: limited SLAs and production-grade support.

Who They Are: AI search engine combining LLMs with real-time web data.

Why Pay Attention: Adds web search for up-to-date, research-oriented summarization.

Capabilities for Summarization: Hybrid LLM + search for current events.

Pricing: Subscription-based.

Data Privacy: Processes queries; review policies.

Scalability & Ecosystem: Best for knowledge workers; limited API.

Business Model & Positioning: Perplexity monetizes through subscription tiers (~$20/month Pro) combining LLM + real-time web search; limited API availability (currently not production-ready for enterprise pipelines). Positioning: specialized for knowledge workers needing web-aware summarization; disadvantage for data analysts: limited API, primarily consumer/knowledge-work focused.

Who They Are: Open-source library for efficient model inference.

Why Pay Attention: Enables cost-effective, self-hosted AI at scale.

Capabilities for Summarization: Optimizes throughput for various models.

Pricing: Free and open-source.

Data Privacy: Runs on your infrastructure.

Scalability & Ecosystem: Supports diverse hardware; high throughput.

Business Model & Positioning: VLLM (open-source) monetizes indirectly through adoption by hosting providers (RunPod, Lambda, AWS Neuron, etc.) who charge for compute; PyTorch Foundation backing supports development. Strengths: highest throughput optimization via PagedAttention, supports all major hardware (NVIDIA, AMD, Intel, TPU, CPU), 74K+ GitHub stars; positioning as foundation for self-hosted, cost-optimized inference at scale.