// the models that matter
AI models
ChatGPT, Claude, Gemini, Grok, Mistral, DeepSeek and more in one place. What each model can do, what it costs, where it is strong and where it falls short. I check the data directly with the vendors and keep it up to date.
Best AI for…
What AI costs
Comparisons
What is coming
All models
Every model in one table. API prices are in US dollars per million tokens, roughly 750,000 words. Click a model for details.
| Model | Vendor | Type | Context window | API price per 1M tokens (input / output) | Open weights | Released |
|---|---|---|---|---|---|---|
| Claude Opus 5.5Anthropic's main model for demanding work, coding and long documents, and the one Anthropic recommends starting with. | Anthropic | flagship | 1M | 4 / 20 | no | Sep 2026 |
| Claude Fable 5.1Claude's most expensive model, meant for the hardest tasks and long autonomous work when even Opus 5.5 falls short. | Anthropic | reasoning | 1M | 10 / 50 | no | Sep 2026 |
| Claude Sonnet 5.5Claude's balanced model for everyday work and coding. It is faster than Opus 5.5 and costs half as much. | Anthropic | balanced | 1M | 2 / 10 | no | Sep 2026 |
| Claude Haiku 5.5Claude's cheapest and fastest model, built for classification, data extraction and large volumes of simple tasks. | Anthropic | fast and cheap | 1M | 0.1 / 0.5 | no | Oct 2026 |
| Command A+An open business model from Canada's Cohere that officially supports Slovak and Czech and runs on your own server. | Cohere | balanced | 128K | – | yes | May 2026 |
| DeepSeek V4 ProDeepSeek's largest open-weight model for demanding text tasks at a low price. It does not understand images. | DeepSeek | flagship | 1M | 1.32 / 3.96 | yes | Aug 2026 |
| DeepSeek V4.1 FlashDeepSeek's cheap, fast open-weight model with a 1 million token window. In the app and API, though, your data goes to China. | DeepSeek | fast and cheap | 1M | 0.3 / 1.2 | yes | Oct 2026 |
| Gemini 4 CarbonGoogle's most capable model, announced on 28 October 2026. So far only selected cybersecurity experts have access. | flagship | – | 2 / 10 | no | Oct 2026 | |
| Gemini 4 ArgonGoogle's most capable publicly available model, out in October 2026. Built for demanding analytical work where accuracy matters and made-up answers are costly. | flagship | 1.05M | 2 / 10 | no | Oct 2026 | |
| Gemini 3.8 FlashGoogle's fast, lower-cost model for everyday work, coding and document analysis. You can try it in the API for free. | balanced | 1.05M | 0.75 / 3.75 | no | Sep 2026 | |
| Gemini 3.5 Flash-LiteA very fast, low-cost Google model for large volumes of simple tasks, such as pulling data out of documents. | fast and cheap | 1.05M | 0.3 / 2.5 | no | Jul 2026 | |
| Gemma 4 31BGoogle's largest open model, which you can run on your own computer or server when data must not leave the company. | small | 256K | – | yes | Mar 2026 | |
| Gemma 4 26B A4BGoogle's faster open model for your own server. Close to the larger Gemma 4 31B in quality, but quicker to respond. | small | 256K | – | yes | Mar 2026 | |
| Muse WonderMeta's strongest model for long-running agents and coding, still released as a preview. | Meta | flagship | 1.05M | 4 / 20 | no | Oct 2026 |
| Muse Spark 1.3Meta's balanced model with a 1-million-token window and a fair price. The Muse Spark line also powers the Meta AI assistant. | Meta | balanced | 1.05M | 1.25 / 4.25 | no | Sep 2026 |
| Muse Spark OpenMeta's large open model that a company can run on its own servers, so data never leaves its infrastructure. | Meta | balanced | 1.05M | 1.25 / 4.25 | yes | Oct 2026 |
| Muse Glimmer 30BMeta's small open model under Apache 2.0 that runs on a single graphics card, even offline. | Meta | small | 131K | 0.3 / 1.2 | yes | Aug 2026 |
| Llama 4 MaverickThe larger model in the newest Llama generation, from April 2025. Cheap through providers, but its license excludes people and companies in the EU. | Meta | balanced | 1M | 0.19 / 0.65 | yes | Apr 2025 |
| Llama 4 ScoutThe smaller Llama 4 from April 2025, very cheap through providers. Its license grants no rights to people or companies in the EU. | Meta | small | 10M | 0.08 / 0.24 | yes | Apr 2025 |
| Mistral Large 4Mistral's largest model, still in preview. Built for demanding tasks, long documents and security analysis with data kept in the EU. | Mistral AI | flagship | 1M | 1.36 / 4.18 | yes | Oct 2026 |
| Mistral Medium 3.5Mistral's mid-size open-weight model and the default in the Vibe app. Good for everyday work and coding with data kept in the EU. | Mistral AI | balanced | 256K | 1.5 / 7.5 | yes | Apr 2026 |
| Mistral Small 4Mistral's cheap, fast open model under Apache 2.0 for everyday tasks through the API or on your own server. | Mistral AI | fast and cheap | 256K | 0.15 / 0.6 | yes | Mar 2026 |
| Ministral 3 14BMistral's small open model under Apache 2.0 that runs on a single GPU, even without an internet connection. | Mistral AI | small | 256K | 0.2 / 0.2 | yes | Dec 2025 |
| Codestral 25.08Mistral's model for code completion right in your editor. Fast and cheap, but dated for autonomous coding agents. | Mistral AI | coding | 128K | 0.3 / 0.9 | no | Jul 2025 |
| Mistral Large 3Mistral's previous large model with open weights under Apache 2.0. Cheap in the API and suited to your own powerful server. | Mistral AI | flagship | 256K | 0.5 / 1.5 | yes | Dec 2025 |
| Kimi K3.1 ProMoonshot AI's newest model from October 2026, built for agentic coding, office documents and very long outputs. | Moonshot AI | flagship | 1.05M | 2 / 10 | no | Oct 2026 |
| Kimi K3Moonshot AI's open model that topped the open-weight rankings in summer 2026, strong at web browsing and computer use. | Moonshot AI | flagship | 1.05M | 3 / 15 | yes | Jul 2026 |
| Nemotron 3.5 Super VLNVIDIA's open model for business AI agents that understands text, images and video and is very fast and cheap to run. | NVIDIA | balanced | 1M | 0.15 / 0.5 | yes | Oct 2026 |
| Nemotron 3.5 Bolt 74B-A9BNVIDIA's open model for AI agents on your own powerful PC, small enough to fit in 24 GB of graphics card memory. | NVIDIA | small | 256K | – | yes | Oct 2026 |
| GPT-6 AstraOpenAI's most capable model, built for the hardest coding, agents that operate a computer and very long documents. | OpenAI | flagship | 1.05M | 10 / 50 | no | Sep 2026 |
| GPT-6.1 SolOpenAI's balanced model, close to GPT-6 Astra in quality at a fifth of the price. A solid pick for business apps. | OpenAI | balanced | 1.05M | 2 / 10 | no | Sep 2026 |
| GPT-6.1 TerraOpenAI's fast, cheaper model for working with images, operating a computer and everyday office work. | OpenAI | fast and cheap | 1.05M | 1 / 5 | no | Oct 2026 |
| GPT-6 LunaOpenAI's cheapest model, made for high volumes of simple tasks, sorting text and low-cost chatbots. | OpenAI | small | 1.05M | 0.1 / 0.5 | no | Sep 2026 |
| gpt-oss-120bOpenAI's open model that a company can run on its own server when data must not leave its infrastructure. | OpenAI | reasoning | 131K | – | yes | Aug 2025 |
| gpt-oss-20bOpenAI's smaller open model that runs on a computer with 16 GB of memory, even without internet access. | OpenAI | small | 131K | – | yes | Aug 2025 |
| Qwen3.8-MaxAlibaba's flagship with a one-million-token context that understands text, images and video. Strong at coding and fairly priced. | Alibaba (Qwen) | flagship | 1M | 2 / 6 | no | Aug 2026 |
| Qwen3.7-PlusAn affordable mid-range Qwen model for everyday business tasks through the API, with a one-million-token context and image input. | Alibaba (Qwen) | balanced | 1M | 0.4 / 1.6 | no | May 2026 |
| Qwen3.8-FlashA very cheap, fast Qwen model for high volumes of tasks. Unlike Max and Plus, it runs in Frankfurt with processing kept in the EU. | Alibaba (Qwen) | fast and cheap | 1M | 0.15 / 0.47 | no | Aug 2026 |
| Qwen3.8-2.4T-A95BThe open-weight text base of Qwen3.8-Max with 2.4 trillion parameters, for large companies with their own GPU cluster. | Alibaba (Qwen) | flagship | 262K | 2 / 6 | yes | Aug 2026 |
| Qwen3.8-27BA small open Qwen model under Apache 2.0 that runs on your own computer and understands images and video. | Alibaba (Qwen) | small | 262K | 0.5 / 3 | yes | Aug 2026 |
| Beam 1.0Reflection AI's large open model under Apache 2.0, built for coding and long-running AI agent tasks. | Reflection AI | flagship | 1M | 0.25 / 1 | yes | Oct 2026 |
| Grok 4.7xAI's strongest model for agentic coding and professional work, with cheaper output than most rivals. | xAI | flagship | 500K | 2 / 6 | no | Sep 2026 |
| Grok 4.3An older, cheaper Grok from April 2026 with a 1-million-token window. Good for lots of low-cost API calls. | xAI | fast and cheap | 1M | 1.25 / 2.5 | no | Apr 2026 |
| Grok Build 0.1A fast Grok coding model and the cheapest Grok on the API, built for agents that write and fix code on their own. | xAI | coding | 256K | 1 / 2 | no | May 2026 |
| GLM-5.3Z.ai's open model for agentic coding, also offered through Mistral's European API with data kept in the EU. | Z.ai (Zhipu AI) | coding | 1M | 1.4 / 4.4 | yes | Aug 2026 |
| GLM-5.3-FlashZ.ai's cheap, fast open model that understands text, images and video and, by Z.ai's account, comes close to top models in tests. | Z.ai (Zhipu AI) | fast and cheap | 1M | 0.15 / 0.5 | yes | Aug 2026 |
Data verified October 28, 2026. Prices and models change often. The data comes from official vendor pricing pages and documentation, and every page shows when it was last checked. Before a larger purchase, confirm the price with the vendor.
Vendors
Anthropic USA
Anthropic is a US company that builds the Claude models and the Claude app for web, desktop and mobile. Its models are also offered through Amazon Bedrock, Google Cloud and Microsoft Foundry. Slovakia and Czechia are both supported countries. The direct API stores data in the US, though, and there is no EU region to choose. Claude Mythos 5.1, which matches Fable 5.1 on specs, is for now limited to verified US organizations.
Claude Opus 5.5, Claude Fable 5.1, Claude Sonnet 5.5, Claude Haiku 5.5 · Claude pricing: what Pro, Max, Team and Enterprise cost
Cohere Canada
Cohere is a Canadian company that builds AI mainly for businesses: the North platform, Compass search, dedicated Model Vault instances and private deployments. Its Command A+ model is fully open under Apache 2.0 and officially supports Slovak and Czech. It does not publish a per-token price for it.
DeepSeek China
DeepSeek is a Chinese company from Hangzhou, known for low-cost models with open weights under the MIT license. Its app, website and API store personal data in China. Czechia's cyber security agency NÚKIB warns against its products, and they are banned for official work in Czech state administration. Open models run on your own server without any link to DeepSeek are a different matter.
Google USA
Google is a US company, and its models are built by the Google DeepMind lab. It offers its closed Gemini models in the Gemini app, in Google Workspace and to developers through the Gemini API. The open Gemma models, under the Apache 2.0 license, can run on your own server. The app and the API work in Slovakia and Czechia, and companies can choose EU data processing on Gemini Enterprise Agent Platform. Its most capable model, Gemini 4 Carbon, announced on 28 October 2026, is so far limited to selected partners.
Gemini 4 Carbon, Gemini 4 Argon, Gemini 3.8 Flash, Gemini 3.5 Flash-Lite, Gemma 4 31B, Gemma 4 26B A4B · Gemini pricing: what Google AI Plus, Pro and Ultra cost
Meta USA
Meta is the US company behind Facebook, Instagram and WhatsApp. Its Meta AI assistant in those apps is free. Its main model line is now Muse, and Llama 4 from April 2025 is still the newest Llama. Some models are released openly, so companies can run them on their own servers. Note that the Llama 4 license excludes people and companies in the EU, while the newer open Muse models have no such restriction.
Muse Wonder, Muse Spark 1.3, Muse Spark Open, Muse Glimmer 30B, Llama 4 Maverick, Llama 4 Scout
Mistral AI France, European vendor
Mistral AI is a French company based in Paris that builds language models. Many of them come with open weights, so a company can run them on its own server. Its app for everyday users was called Le Chat until May 2026 and is now Vibe. Mistral stores data in the EU by default, which makes GDPR compliance easier for European companies.
Mistral Large 4, Mistral Medium 3.5, Mistral Small 4, Ministral 3 14B, Codestral 25.08, Mistral Large 3 · Mistral Vibe (Le Chat) pricing: what Pro and Team cost
Moonshot AI China
Moonshot AI is a Chinese company that develops the Kimi models and runs Kimi.com, the Kimi app, Kimi Work and Kimi Code. Its Kimi K3 was the top-rated open model on the Artificial Analysis index in August 2026. Its international API bills in USD, and its privacy policy says data may be stored on servers in Singapore.
NVIDIA USA
NVIDIA is a US company based in Santa Clara, California, best known for graphics cards and AI chips. It also builds the open Nemotron language models and publishes their weights, open datasets and training recipes. It has no paid chat subscription for everyday users. Companies download the models and run them on their own hardware or use them through API providers.
OpenAI USA
OpenAI is the US company that launched ChatGPT in 2022 and kicked off the wave of AI chatbots. Besides ChatGPT, it sells its GPT models to developers through an API and has released the open gpt-oss models. ChatGPT and the API work in Slovakia and Czechia. Companies on Enterprise and API customers can choose to have their data processed in the EU.
GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Luna, gpt-oss-120b, gpt-oss-20b · ChatGPT pricing: what Go, Plus, Pro and Business cost
Alibaba (Qwen) China
Qwen is a family of AI models from the Qwen team at China's Alibaba Group. Alibaba Cloud sells them through its Model Studio service, and many are also released as open weights you can download and run yourself. Two API regions matter for Europe: Singapore and Frankfurt. Of the main models, though, Frankfurt only offers EU-bound processing for Qwen3.8-Flash.
Qwen3.8-Max, Qwen3.7-Plus, Qwen3.8-Flash, Qwen3.8-2.4T-A95B, Qwen3.8-27B
Reflection AI USA
Reflection AI is a US company that, according to TechCrunch, was founded in 2024 by two former Google DeepMind researchers and is based in Brooklyn. PitchBook data puts its funding at about $4.7 billion, with backers including NVIDIA, Sequoia Capital and Lightspeed. The company says it wants to build open AI that developers, businesses and governments can control. Beam, released in October 2026, is its first open model. It has no chat app with a subscription for everyday users.
xAI USA
xAI is the US company founded by Elon Musk that builds the Grok chatbot and the Grok models for developers. Since February 2026 it has belonged to SpaceX and its website now uses the SpaceXSI brand, while the Grok, Grok Build and API names stay the same. The newest model is Grok 4.7 from September 2026, while the cheaper Grok 4.3 dates from April. For European companies the key point is that xAI only guarantees where data is processed in the US. It has no European endpoint or EU data zone.
Grok 4.7, Grok 4.3, Grok Build 0.1 · Grok pricing: what Plus, Premium, Super and Ultra cost
Z.ai (Zhipu AI) China
Z.ai (Zhipu AI) is a Chinese company that develops the open GLM models. It sells them through its own API in USD and through the GLM Coding Plan subscription for coding tools. What matters for Europe is that France's Mistral AI also offers the GLM-5.3 models in its API, where data is stored in the EU by default.
How this catalog is built
The catalog covers current models that matter to businesses and everyday users in Europe. Specs and prices come from official vendor pages, quality ratings from independent benchmarks and from my own work, since I use these models every day on client projects. Every figure links to its source.
Not sure which model fits your business?
I can help you pick a model and plan based on what you actually need, set it up safely for company data and train your team to use it well. Let’s start with an intro consultation.