AI Models Hub | Madriguera de Datos

//AI Models & Big Tech Hub

Strategic guide to models from the big AI companies

The big companies & their AI models

An interactive map to understand each AI's architecture, when to use it, what it's for, and how to integrate it into your data pipelines.

Not sure which model to pick? Quick selection assistant:

⚡

OpenAI

Creators of ChatGPT and the GPT family. In September 2026 they launched GPT-6 in three sizes: Astra, Sol and Luna.

General purposeContext: See OpenAI docs

GPT-6.1 Sol$2 / $10 per M tokens

The balanced model of the GPT-6 family (launched September 22, 2026, now at version 6.1): strong reasoning and speed at half the price of its GPT-5.6 equivalent. Since October 8 it has an Ultrafast mode up to 8x faster, at $12 / $60 per M tokens.

Speed★★★★☆

Reasoning★★★★★

Cost$2 / $10 per M tokens

Ideal use cases:

Internal assistants and customer service with high answer quality

Generating Python, SQL and DAX

Analyzing documents, tables and images

Executive writing and report summaries

Brand: OpenAI

Volume & low costContext: See OpenAI docs

GPT-6 Luna$0.10 / $0.50 per M tokens

The budget GPT-6 model, built for bulk, repetitive tasks where cost per token rules.

Speed★★★★★

Reasoning★★★☆☆

Cost$0.10 / $0.50 per M tokens

Ideal use cases:

Classifying thousands of surveys, complaints or tickets

Extracting fields from invoices, forms and emails

Routing requests to the right model or team

Tagging and cleaning text data

Brand: OpenAI

🧠

Anthropic

Creators of Claude. Between September and October 2026 they refreshed all three tiers: Opus 5.5, Sonnet 5.5 and Haiku 5.5.

Reasoning & agentsContext: See Anthropic docs

Claude Opus 5.5$4 / $20 per M tokens

Claude’s large model (September 22, 2026). It matches Claude Fable 5.1 on most tasks and costs 40% less to run than Opus 5. It ships with adaptive thinking.

Speed★★★☆☆

Reasoning★★★★★

Cost$4 / $20 per M tokens

Ideal use cases:

Agents that plan and carry out multi-step tasks

Analyzing long documents: regulations, contracts, reports

Complex programming and architecture review

Decisions where a mistake costs more than the model

Brand: Anthropic

Code & everyday workContext: See Anthropic docs

Claude Sonnet 5.5$2 / $10 per M tokens

Claude’s middle tier (September 28, 2026): 30% faster and up to 30% cheaper than Sonnet 5. The default model for most workflows.

Speed★★★★☆

Reasoning★★★★★

Cost$2 / $10 per M tokens

Ideal use cases:

Building apps, data scripts and dashboards

Writing and editing reports and articles

Service agents and internal assistants

n8n automations that need judgment

Brand: Anthropic

Fast & affordableContext: See Anthropic docs

Claude Haiku 5.5$0.10 / $0.50 per M tokens (prompts up to 100K)

Claude’s small model (October 7, 2026), presented as the fastest, cheapest and most capable of its size. Built for high volume.

Speed★★★★★

Reasoning★★★★☆

Cost$0.10 / $0.50 per M tokens (prompts up to 100K)

Ideal use cases:

Sub-agents that classify, summarize or route

Extracting data from documents at scale

Low-latency customer service chat

Pre-filters before calling a large model

Brand: Anthropic

✨

Google DeepMind

Creators of Gemini, built into Google Workspace, Android and now Windows. In September 2026 they unveiled Gemini 4 Argon and the Gemini 3.8 family.

Frontier · available to everyoneContext: Up to 1M tokens of input and output

Gemini 4 Argon$2 / $10 per M tokens until Nov 30; then $4 / $20

Google’s most capable model. Announced on September 30 for cybersecurity defenders, it has been open to everyone since October 13, 2026: Gemini app (Pro and Ultra), AI Studio, API and Gemini Enterprise.

Speed★★★☆☆

Reasoning★★★★★

Cost$2 / $10 per M tokens until Nov 30; then $4 / $20

Ideal use cases:

Financial and legal analysis

Software engineering and code refactoring

Generating very long documents and code

Brand: Google DeepMind

Fast & affordableContext: See Google AI docs

Gemini 3.8 Flash$0.75 / $3.75 per M tokens (doubles Jan 1, 2027)

Google’s fast model (September 2, 2026): better reasoning and coding at lower cost. Its promotional price runs until December 31, 2026.

Speed★★★★★

Reasoning★★★★☆

Cost$0.75 / $3.75 per M tokens (doubles Jan 1, 2027)

Ideal use cases:

Document summaries and analysis in Google Workspace

High-frequency RAG flows

Entity extraction and JSON outputs in n8n

Voice agents with Gemini 3.8 Live

Brand: Google DeepMind

🐋

DeepSeek

Chinese lab behind very low-cost open models. In October 2026 it was closing a round of at least $12 billion and preparing a Hong Kong IPO.

Open reasoningContext: See DeepSeek docs

DeepSeek ReasonerVery low / open weights

DeepSeek’s reasoning line, with open weights and an API compatible with the OpenAI SDK. The API alias always points to the current version.

Speed★★★☆☆

Reasoning★★★★☆

CostVery low / open weights

Ideal use cases:

Local deployment (vLLM / Ollama) when data cannot leave

Low-cost math and logic reasoning

Distilling smaller models

Benchmarking against closed models

Brand: DeepSeek

♾️

Meta AI

Meta moved from the Llama family to Muse Spark, its model line for Meta AI and the API.

Agents with fewer callsContext: See Meta docs

Muse Spark 1.3$1.25 / $4.25 per M tokens

The September 2026 release of Muse Spark: it makes about 20% fewer tool calls than 1.2 to solve the same task. A much cheaper “Contributor” variant trains on user traffic.

Speed★★★★☆

Reasoning★★★★☆

Cost$1.25 / $4.25 per M tokens

Ideal use cases:

Agents that use many tools, where every call costs

Assistants integrated with WhatsApp, Instagram or Messenger

Cost tests against other mid-size models

Brand: Meta AI

🌬️

Mistral AI

French lab building open-weight models, the leading European option for anyone who needs data sovereignty.

Open frontier · previewContext: 1,000,000 tokens

Mistral Large 4API preview; open weights at the end of October

Mistral’s largest model (October 6, 2026): a mixture of experts with over a trillion parameters, about 49 billion active, image input and a one-million-token context. Trained in its own EU data centers.

Speed★★★☆☆

Reasoning★★★★★

CostAPI preview; open weights at the end of October

Ideal use cases:

Regulated sectors that need to host the model on their own infrastructure

Analyzing full regulations, contracts and document bases

Code, agents and multilingual tasks

Fine-tuning on your own data once weights ship

Brand: Mistral AI

🛰️

xAI

Elon Musk’s AI company, maker of Grok, built into X.

Coding & agentsContext: See xAI docs

Grok 4.7$2 / $6 per M tokens

The September 2026 release of Grok, on a larger base model than Grok 4.6. Its benchmark results, for now, have only been published by xAI.

Speed★★★★☆

Reasoning★★★★☆

Cost$2 / $6 per M tokens

Ideal use cases:

Programming and code review

Agentic flows where output cost matters

Cost and quality comparisons against Sonnet and Sol

Brand: xAI