Home › LLM Landscape

The LLM Landscape for Business

Last updated: 2026-07-21 · Updated quarterly · Maintained by Shubin Yu

The working map of major AI model families for business buyers: who makes them, how you buy them, and what each is for. Families and positioning are durable; point versions churn monthly and belong in your quarterly check, not your memory.

Major model families and their business positioning (2026-07-21)
ProviderModel familyAccessBusiness positioning
AnthropicClaude (Fable / Opus / Sonnet / Haiku)API, AWS Bedrock, Google Cloud, AzureFrontier flagship tier plus workhorse tiers; strength in agentic coding, long-horizon autonomy, enterprise reliability; leading enterprise API share
OpenAIGPT (flagship / mid / small tiers)API, AzureFrontier generalist family; strongest consumer footprint; deep agent platform (Responses API, Agents SDK, Codex)
GoogleGemini (Pro / Flash)API, Vertex AI, WorkspaceMultimodal-first; huge context windows; strongest suite integration (Workspace, NotebookLM)
Moonshot AIKimi (K-series)API, open weights, hostsOpen-weight frontier challenger; near-flagship capability, self-hostable, aggressive pricing
DeepSeekDeepSeek (V-series / R-series)API, open weights (MIT), hostsUltra-low-cost open-weight line; efficiency-first architectures; strong coding and reasoning
AlibabaQwenAPI, open weights, hostsDominant small-model tier that runs on ordinary hardware; broad size range
Zhipu AIGLMAPI, open weights (MIT)Open-weight frontier line from China; strong reasoning benchmarks
xAIGrokAPI, X platformReal-time X data access; repositioned toward coding and agents
MetaLlamaOpen weights, hostsThe Western open-weight standard-bearer; broad ecosystem, slower recent cadence
Mistral AIMistral (Large / Small)API, open weights (Apache)European champion; open-weight flagships; EU sovereignty appeal

How to read this table

Model point-versions churn monthly; families and positioning are durable. For live benchmarks and prices, check LMArena and Artificial Analysis, then the provider pricing pages. The selection method — capability, cost, and sovereignty, not brand familiarity — is covered in Chapter 4 of the book, and the quarterly fifteen-minute check-up habit in its Section 4.6.

Changes this quarter

2026-07: Page created from the book's mid-2026 snapshot. Open-weight challengers (Kimi, DeepSeek, GLM) now sit within striking distance of the closed frontier; enterprise procurement continues shifting toward cloud marketplaces (Bedrock, Azure, Vertex) for data-residency reasons.

This page implements the habit taught in GenAI for Business: series names in your mental model, versions checked quarterly, decisions based on your own evals.