MODEL INTELLIGENCE
Enterprise Model Market Map — Hosted Platforms vs Open-Weight
A pragmatic lens on procurement, security, latency, cost, and governance tradeoffs. What changes matter right now, which platforms are production-ready, and when to self-host vs use hosted APIs.
Updated
January 2025 — reflects latest model releases and pricing
Coverage
OpenAI, Anthropic, Google, Meta, Mistral, Cohere + deployment tradeoffs
Use this for
Vendor selection, cost modeling, compliance planning, PoC scoping
Platform Overview
Major platforms as of January 2025, with models, strengths, and procurement readiness.
OpenAI
Hosted API
Enterprise-Ready
Key Models
GPT-4oGPT-4o-minio1o1-mini
Strengths
- Market-leading reasoning and coding (o1)
- Mature API, extensive tooling ecosystem
- Azure OpenAI for enterprise compliance (BAA, FedRAMP)
Considerations
- Data retention policies (check Azure vs API)
- Rate limits and quota management for production
- Cost optimization via prompt caching and batch API
Anthropic
Hosted API
Enterprise-Ready
Key Models
Claude 3.5 SonnetClaude 3 OpusClaude 3 Haiku
Strengths
- 200K context window for long-document analysis
- Strong refusal training, excellent for safety-critical apps
- AWS Bedrock integration for compliance
Considerations
- Token limits on output (4K-8K max per response)
- Pricing: Opus expensive for high-volume workloads
- AWS Bedrock required for FedRAMP/HIPAA
Google (Gemini)
Hosted API
Enterprise-Ready
Key Models
Gemini 1.5 ProGemini 1.5 FlashGemini 2.0
Strengths
- Multimodal (text, image, video, audio) in single API
- 1M+ token context window for massive documents
- Vertex AI for GCP-native enterprise deployments
Considerations
- Gemini 2.0 (experimental) - check production readiness
- GCP Vertex AI pricing differs from public API
- Data residency and compliance via Vertex AI
Meta (Llama)
Open-Weight
Self-Serve
Key Models
Llama 3.3 (70B)Llama 3.1 (8B, 70B, 405B)Llama Guard
Strengths
- Run on-prem or in your VPC (full data control)
- No per-token API costs—fixed infrastructure cost
- Llama Guard for content moderation, safety filtering
Considerations
- Requires ML infrastructure (GPU clusters, inference servers)
- Model fine-tuning expertise for domain-specific tasks
- Latency and throughput tuning vs hosted APIs
Mistral AI
Hybrid
Mixed
Key Models
Mistral LargeMistral 7BMixtral 8x7B
Strengths
- Hosted API + open-weight models (flexibility)
- Strong multilingual support (French, Spanish, etc.)
- Azure integration for enterprise compliance
Considerations
- Smaller context windows (32K-128K)
- Less mature ecosystem vs OpenAI/Anthropic
- Open-weight models require self-hosting expertise
Cohere
Hosted API
Enterprise-Ready
Key Models
Command R+Command REmbed v3
Strengths
- Enterprise RAG focus (retrieval, reranking, embeddings)
- Multi-turn conversational agents with citations
- Deployment flexibility (API, AWS, GCP, Azure, on-prem)
Considerations
- Best for RAG use cases—less general-purpose
- Pricing model differs from token-based APIs
- Check model availability in your preferred cloud
Tradeoff Matrix: Hosted API vs Open-Weight
Key decision dimensions for selecting deployment approach. No universal answer—tradeoffs depend on your constraints.
Data Control
Hosted API
Data sent to vendor (check DPA, BAA, retention policies)
Open-Weight
Full control—run in your VPC or on-prem
Recommendation
Use open-weight for PII/PHI; hosted APIs for non-sensitive use cases
Cost Structure
Hosted API
Per-token pricing (scales with usage, unpredictable at high volume)
Open-Weight
Fixed infra cost (GPU clusters, ML engineers, but predictable)
Recommendation
Hosted APIs for MVP/low-volume; open-weight at scale (1M+ requests/day)
Latency
Hosted API
Network RTT + queue time (100-500ms typical)
Open-Weight
Controlled by your infra (can optimize to <50ms p95)
Recommendation
Hosted APIs for async workflows; open-weight for real-time UX
Compliance
Hosted API
Vendor compliance (SOC 2, FedRAMP, HIPAA via cloud integrations)
Open-Weight
Your compliance posture—vendor only provides model weights
Recommendation
Use cloud-integrated APIs (Azure OpenAI, AWS Bedrock) for regulated industries
Model Updates
Hosted API
Automatic updates (breaking changes possible, version pinning recommended)
Open-Weight
You control versions (manual updates, but stable)
Recommendation
Pin versions in production; test updates in staging first
Want a vendor selection tailored to your use case?
We deliver model selection guidance customized to your requirements: cost envelope, latency targets, data residency, compliance needs, and existing cloud contracts. Output includes vendor shortlist, cost modeling, PoC plan, and procurement-ready RFP criteria.