Map of major players, model families, and the open source vs closed source debate.
The AI landscape changes too quickly for a static list of model numbers. This module gives you a current map of the leading labs, how their model families fit together, and a durable way to choose between them.
No single ranking captures the whole market. These selected labs are influential across frontier capability, open-weight releases, developer adoption, and product reach.
GPT-5.6 Sol · GPT-5.6 Terra · GPT-5.6 Luna
The GPT-5.6 family now uses named tiers: Sol is the flagship for the hardest work, Terra balances capability and throughput, and Luna is the efficient tier for speed-sensitive workloads.
Key strength: Agentic coding, long-context knowledge work, multimodality, and a broad product ecosystem.
Official model information ↗Claude Fable 5 · Mythos 5 (limited) · Opus 5 · Sonnet 5 · Haiku 4.5
Anthropic has expanded Claude beyond the familiar Opus, Sonnet, and Haiku ladder. Fable 5 is its top broadly available model, while Mythos 5 is a restricted research deployment.
Key strength: Long-running agents, coding and analysis, safety research, and the Model Context Protocol ecosystem.
Official model information ↗Gemini 3.6 Flash · 3.5 Flash · 3.5 Flash-Lite · 3.1 Pro (preview)
Gemini spans experimental Pro models, high-capability Flash models, and lower-cost Flash-Lite models. The family is multimodal and tightly connected to Google products and developer tooling.
Key strength: Multimodal systems, scientific research, broad infrastructure, and on-device deployment.
Official model information ↗Grok 4.5
Grok 4.5 is xAI's current flagship model for agentic software work and complex workflows, with native text and image input.
Key strength: Large-scale compute, software engineering workflows, and integration with the X ecosystem.
Official model information ↗Kimi K3 · Kimi K2.6
Kimi K3 is Moonshot's 2.8-trillion-parameter flagship for long-horizon coding, knowledge work, and reasoning. It supports native vision and up to a one-million-token context window.
Key strength: Long-context agentic work, coding, native vision, and a fast-moving open-model programme.
Official model information ↗GLM-5.2
GLM-5.2 is Z.ai's current flagship for long-horizon agent tasks. It combines a one-million-token context window with an MIT-licensed open-weight release.
Key strength: Open frontier models, agentic coding, long context, and multilingual capability.
Official model information ↗Mistral Medium 3.5 · Small 4 · Large 3 · Devstral 2
Mistral combines efficient general-purpose models with specialised coding and reasoning families. Its catalogue mixes downloadable models with managed commercial offerings.
Key strength: Efficient architectures, multilingual performance, specialised models, and European deployment options.
Official model information ↗Llama 4 Scout · Llama 4 Maverick
Meta's Llama family remains a major foundation for self-hosted and community-built AI. Llama 4 introduced natively multimodal mixture-of-experts models.
Key strength: Open-weight distribution, research, community adoption, and a broad deployment ecosystem.
Official model information ↗Model names differ by vendor, but most families still serve four practical needs. Choose the least expensive tier that reliably completes your task, then evaluate with your own examples.
| Tier | Purpose | Current examples | Typical use |
|---|---|---|---|
| Frontier | Maximum capability for the hardest work | GPT-5.6 Sol, Claude Fable 5, Kimi K3 | Long-running agents, research, complex engineering |
| High capability | Deep reasoning with production flexibility | GPT-5.6 Terra, Claude Opus 5, Gemini 3.1 Pro, Grok 4.5, GLM-5.2 | Analysis, coding, planning, enterprise workflows |
| Balanced | Strong quality at useful speed and cost | Claude Sonnet 5, Gemini 3.6 Flash, Mistral Medium 3.5 | Everyday knowledge work and production applications |
| Fast / light | Low latency and high throughput | GPT-5.6 Luna, Claude Haiku 4.5, Gemini 3.5 Flash-Lite, Mistral Small 4 | Classification, extraction, routing, high-volume tasks |
"Open source" is often used loosely in AI. A more precise distinction is whether you can download the model weights or must access the model through a provider.
Llama 4, GLM-5.2, Mistral models, Qwen, and Kimi K3 (full weights announced for 27 July 2026)
GPT-5.6, Claude 5, Gemini 3, and Grok 4.5
Foundation models are one layer in a much larger stack. Real-world systems also depend on compute, orchestration, evaluation, data infrastructure, and user-facing applications.
OpenAI Developers
Official guide to the GPT-5.6 Sol, Terra, and Luna family.
Anthropic
Current Claude model family and official capability guidance.
Google AI for Developers
Official index of stable, preview, and experimental Gemini models.
Moonshot AI
Official release overview for Kimi K3 and its architecture.
Rowan Cheung
A daily overview of AI product launches and industry moves.
Nathan Benaich & Air Street Capital
An annual report covering AI research, industry, policy, and safety.
7 questions · Score 70% or higher to complete this module
You can retake the quiz as many times as you need. Your best score is saved.
