Leading Large Language Models
Discover the most powerful AI models and find the perfect solution for your business. All models can be integrated into CompanyGPT.
Which AI Model Fits Your Business?
Choosing the right AI model is crucial for the success of your AI strategy. Each model has its strengths: GPT-6 Astra and GPT-6.1 Sol for reasoning, coding and computer use, Claude Opus 5.5 for safety and long context windows, Gemini 4 Argon and 3.8 Flash for multimodal tasks and long workflows. Open-weight models such as Qwen 3.8, GLM-5.3, DeepSeek V4.1 or Kolibri-1 offer maximum control and data sovereignty. New as a category of its own: decision models (System One) such as Jev, Clef and Laya, which do not generate text but return typed decisions with probabilities.
The model thinks, the harness acts: AI harnesses such as Claude Code, Codex and Pi at a glance →
Premium AI Models
The most powerful models for enterprise applications
GPT-6.1 Sol & Terra, GPT-6 Astra
OpenAIThe sixth generation: Astra as the top model, GPT-6.1 Sol close to Astra per OpenAI, the new GPT-6.1 Terra fast and low-cost for image analysis and computer use, Luna for volume – all four tiers in the EU Data Zone via Microsoft Foundry.
Claude Opus, Sonnet & Haiku 5.5
AnthropicAnthropic's Claude 5.5 family: Opus for complex agentic work, Sonnet for everyday use, Haiku for high volume – without mandatory 30-day retention, in the EU via Google Cloud and AWS Bedrock.
Gemini 4 Argon
GoogleFirst model of the Gemini 4 generation for coding, analysis and long agent workflows, with data residency in the EU multi-region; plus Gemini 3.8 Flash as the default model.
Open Source Models
Full control and maximum data sovereignty
Llama, Muse Glimmer & Muse Spark Open
MetaMuse Glimmer 30B under Apache 2.0 for self-hosting in the EU, Muse Spark Open with open weights under the Muse licence, Muse Wonder as an API preview; Llama 3.3 also via STACKIT.
Mistral Large 4, Medium 3.5 & Small 4
Mistral AI (France)Mistral Large 4 (1M context) in public preview, weights announced; Medium 3.5, Small 4 and Large 3 (Apache 2.0) remain the production picks. New: Voxtral 3 for speech.
DeepSeek V4.1 Flash
DeepSeek (China)Open MoE models of the V4 generation under the MIT licence. Self-hosting in the EU recommended, the official API runs on Chinese infrastructure.
Qwen 3.8
Alibaba (China)Qwen3.8-27B under Apache 2.0 as the self-hosting favourite, Qwen3.8-Flash-Next as an open-weight preview of Qwen 4; API also via Alibaba Cloud Frankfurt.
GLM-5.3
Zhipu AI / Z.ai (China)Powerful open-weight model with strong coding and agentic performance. Self-hosting possible.
MiniMax-M3.1 Flash & M3
MiniMax AI (China)Open-weight models with 1M context and image and video input for agentic coding: M3.1 Flash (428B parameters, 23B active) as the efficient variant, M3 as the larger model; community licence with attribution requirement, self-hosting possible.
Kolibri-1
Aleph Alpha (Germany)German-English MoE model (78.1B parameters, 3.46B active, 1M context) under Apache 2.0 – trained in Germany and Finland, self-hosting from two H100s.
FLUX 3
Black Forest Labs (Germany)Image generation from Germany: FLUX 3 as the current generation, open variants such as FLUX.2 [klein] 4B (Apache 2.0) and FLUX 3 Image Klein for self-hosting.
Kimi K3.1 Pro & K3
Moonshot AI (China)Kimi K3.1 Pro via API, weights announced; K3 and K2.7 Code with open weights for self-hosting in the EU – pricing, licence and GDPR routes at a glance.
Decision Models (System One)
No text, but a typed decision with a probability: classify, route, score – as a pre-stage or control instance for language models and agents.
Jev
TypeSafe AI (USA)The model that shaped the category: question types noul, choice and score, 64k context, API only from the US.
Clef, Clef-omni & Clef-flash
Cloudflare (USA)Open, multimodal decision models (27B, 30B-A3B and 9B, Apache 2.0), Jev-API-compatible, Clef-omni also with audio – on Workers AI from USD 0.038 per 1M input tokens or self-hosted in the EU.
Laya
Convai InnovationsEncoder-based decision model with 0.4B parameters under Apache 2.0 – runs on CPU, self-hosting only.
Julia-1
Supersonic Labs144M parameters on an mmBERT base, Jev-compatible question types, 52 languages – Apache 2.0, runs on CPU.
GLiNER2.5-Decide
Fastino340M parameters on a DeBERTa base for schema-defined decisions (intent routing, triage) – Apache 2.0, CPU-capable; multilingual variant separate.
Microsoft-Decision-1
MicrosoftCalibrated probabilities for yes/no, choice and rating – USD 0.042 per 1M input tokens, output free, in Microsoft Foundry also as an EU Data Zone deployment.
OpenAI Decisions API
OpenAIGenerally available since 22 October 2026: GPT-6 Luna answers up to 400 questions per request with probabilities – USD 0.10 per 1M input tokens, output free.
Specialized Models
Optimized for specific use cases
Google Veo 3.1
GoogleGoogle's video model with native audio output via Vertex AI – processing in the US only per Google's documentation.
LTX-2.5
LightricksOpen video and audio model with 22B parameters, up to 4K, open weights for self-hosting in the EU (community licence with revenue threshold).
Kling 4.0
Kuaishou (China)Video generation from Kuaishou: Kling 4.0 with clips up to 30 seconds and stereo audio. Servers in Singapore per the privacy policy, no EU region documented.
Suno v6
Suno (USA)AI music generator with v6, v6-wild and v6-mini: songs from text and audio, Studio 2.0, commercial use on paid plans only – legal situation and GDPR at a glance.
gpt-transcribe & Whisper
OpenAITranscription in the OpenAI API with gpt-transcribe and gpt-live-transcribe; Whisper remains an open model for self-hosting, whisper-1 in the API ends on 26 February 2027.
Mistral OCR 4.1
Mistral AI (France)Document processing and OCR with bounding boxes and block confidence, generally available. EU provider from France.
MedGemma
GoogleOpen medical models for research: MedGemma 1.5 4B for imaging including CT and MRI, MedGemma 27B for complex text tasks.
xAI Grok 4.7
xAI (Elon Musk)Grok 4.7 and 4.6 from xAI with up to 500k context and reasoning levels; data protection and EU options in detail on the model page.
All AI Models in Detail
Click on a model for detailed information
AI21 Jamba
AI21 LabsAI21 Jamba2: Mamba-Transformer hybrid with 256k context, open weights under Apache 2.0, EU …
Aleph Alpha (Kolibri-1, PhariaAI, Luminous)
Aleph AlphaAleph Alpha from Heidelberg: Kolibri-1 (Apache 2.0, MoE), PhariaAI, Pharia-1 and T-Free …
Aleph Alpha Kolibri-1
Aleph AlphaKolibri-1 by Aleph Alpha: open German-English MoE model (78.1B parameters, 1M tokens, Apache 2.0) …
Alibaba Qwen
AlibabaAlibaba Qwen 3.8: Qwen3.8-27B (Apache 2.0), Flash-Next, Max and Omni-Flash. Licences, API pricing, …
Amazon Titan & Nova
Amazon Web ServicesAmazon Nova and Titan on AWS Bedrock: Nova 2 Lite, Nova 2.5 Sonic, Pro and Omni as the current …
Anthropic Claude
AnthropicAnthropic Claude overview: Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1 with pricing, GDPR, DPA, EU …
Black Forest Labs FLUX
Black Forest LabsFLUX by Black Forest Labs (Freiburg): FLUX.2 Max, Pro, Flex, Dev, Klein and FLUX 3 for image and …
ByteDance Seedance 2.0
ByteDanceSeedance 2.0 and 2.5 by ByteDance: audio-video model with synchronised sound and multi-shot …
Cloudflare Clef (Clef & Clef-flash)
CloudflareCloudflare Clef, Clef-omni and Clef-flash: open decision models (Apache 2.0) with probabilities …
Cohere Command
CohereCohere Command A+ (Apache 2.0, MoE), Command A, Reasoning and Vision: enterprise LLMs with RAG …
DeepSeek
DeepSeekDeepSeek V4.1-Flash-1017 and V4-Pro: 1M context, MIT weights, peak/off-peak API pricing, …
ElevenLabs
ElevenLabsElevenLabs Eleven v4, v4 Turbo, Eleven Realtime, Scribe v2, Music v2.5 and Dubbing v2: models, …
GLiNER2.5-Decide (Fastino)
FastinoGLiNER2.5-Decide by Fastino: open decision model (340M parameters, Apache 2.0) answering with …
GLM-5.3
Zhipu AIGLM-5.3 and GLM-5.3-Flash by Z.ai: open weights, licences, benchmarks, API pricing, Bedrock …
Google Gemini
GoogleGoogle Gemini 4 Argon and 3.8 Flash: versions, prices, benchmarks, EU multi-region with DRZ and MLP …
Google Gemma
GoogleGoogle Gemma 4 and EmbeddingGemma 2: open-weights models under Apache 2.0 from E2B to 31B plus …
Google MedGemma
GoogleGoogle MedGemma: open models for medical image and text analysis, MedGemma 1.5 4B with CT, MRI and …
Google Veo
GoogleGoogle Veo 3.1 (Standard, Fast, Lite) and Gemini Omni 1.1 Flash: pricing, availability and …
Jev (TypeSafe AI)
TypeSafe AIJev by TypeSafe AI is not an LLM but a System One model: typed answers with probabilities, USD 0.042 …
Krea
Krea AIKrea AI (USA): generative-media suite for image, video, realtime canvas and upscaling. Krea 3, Krea …
Kuaishou Kling
KuaishouKling 4.0 and 3.0 by Kuaishou: video generation with audio, clips up to 30 seconds. Access via app, …
Laya (Convai Innovations)
Convai InnovationsLaya by Convai Innovations: open decision model (Apache 2.0, 322 to 421M parameters) with a …
Lightricks LTX-2.5
LightricksLTX-2.5 by Lightricks: open video-audio model with 22B parameters, up to 4K with sound, native …
Meta Llama & Muse
MetaMeta Muse and Llama: Muse Wonder (preview, $4/$20), Muse Spark Open 816B (open weights), Muse …
Microsoft Phi & MAI
MicrosoftMicrosoft MAI and Phi: MAI-Transcribe-2 and Streaming, MAI-Voice-2.1, MAI-Thinking-1, …
Microsoft-Decision-1
MicrosoftMicrosoft-Decision-1: decision model in Microsoft Foundry with calibrated probabilities per answer …
MiniMax-M3
MiniMax AIMiniMax-M3.1-Flash and MiniMax-M3: open-weight models with 1M context for agentic coding. API …
Mistral AI
Mistral AIMistral AI: Large 4 (preview, 1T parameters, 1M context), Medium 3.5, Small 4, Voxtral 3 and OCR …
Moonshot Kimi
Moonshot AIMoonshot Kimi K3.1 Pro, K3, K2.7 Code and K2.6: 1M context, API pricing, licence, weights, …
Nano Banana (Gemini 3 Pro Image)
GoogleNano Banana Pro (Gemini 3 Pro Image), 2.1, 2 and 2 Lite compared: pricing, status, regions. Only …
NVIDIA Nemotron
NVIDIANVIDIA Nemotron 3.5 (Super VL, Bolt, Lightning) and Nemotron 3 (Nano, Super, Ultra): open models for …
OpenAI GPT
OpenAIOpenAI GPT: GPT-6.1 Terra, GPT-6.1 Sol, GPT-6 Luna and Astra at a glance – pricing, EU Data Zone in …
OpenAI gpt-oss
OpenAIgpt-oss-120b and gpt-oss-20b: OpenAI open-weight models (Apache 2.0) for GDPR-compliant self-hosting …
OpenAI Privacy Filter
OpenAIOpenAI Privacy Filter: open-source model (Apache 2.0, 1.5B parameters) for detecting and redacting …
OpenAI Whisper
OpenAIOpenAI Whisper: open-source speech-to-text (MIT, large-v3-turbo) for GDPR-compliant transcription …
OpenEuroLLM
OpenEuroLLM ConsortiumOpenEuroLLM: EU-funded open-source LLM project with 20 partners. First base model Prelude 9B (Apache …
SOOFI (Soofi S)
SOOFI Consortium (German AI Association)Soofi S (SOOFI): sovereign, open-source 30B AI language model (LLM) from Germany – MoE with 3B …
Stability AI
Stability AIStability AI: Stable Diffusion 3.5, Stable Audio 3.0 and Stable Video Diffusion. Open-weights image …
Suno
SunoSuno v6, v6-wild and v6-mini: AI music generator with Studio 2.0, pricing, usage rights, label …
Supersonic Labs Julia-1
Supersonic LabsJulia-1 by Supersonic Labs: open decision model (System One, 144.3M parameters, Apache 2.0) for …
Tencent Hunyuan (Hy3)
TencentTencent Hunyuan Hy3 and Hy4-Preview: open-source MoE models under Apache 2.0 without EU …
xAI Grok
xAIxAI Grok: Grok 4.7, Grok 4.6 and Grok Build 0.1 at a glance – API pricing, 500k context, reasoning …
GDPR-Compliant Cloud Infrastructure
All models in CompanyGPT run on European infrastructure - your data never leaves the EU.
Microsoft Azure
OpenAI models (GPT-6.1 Sol and Terra, GPT-6 Astra, Sol and Luna, GPT-5.6) in Microsoft Foundry as Data Zone Standard (EU) – processing within the EU, including Germany West Central. Enterprise-grade security, SOC 2 & ISO 27001 certified.
Google Vertex AI
Gemini 4 Argon, Gemini 3.8 Flash and Claude models on Google Cloud – Gemini with EU data residency, Claude in the EU multi-region or in europe-west1 (Belgium). Multimodal capabilities with enterprise-grade security and compliance.
AWS Bedrock
Claude, Mistral, gpt-oss and more models on Amazon Bedrock – Claude in the EU via eu. geo inference profiles, selected open models in-region in Frankfurt. Seamless integration into existing AWS infrastructure with IAM and VPC.
STACKIT Model Serving
German sovereign cloud by Schwarz Digits. Open models such as gpt-oss-120b, Llama 3.3, Gemma 4 and Qwen 3.8 via an OpenAI-compatible API in region eu01 on German infrastructure.
You can source all four clouds directly through innFactory GmbH as a reseller: billed by invoice instead of credit card, all providers on one invoice, with a personal contact. Cloud procurement via innFactory →
Local or Cloud? Transparent Cost Comparison
Is your own GPU server worth it for open-weight models like Qwen, Mistral, DeepSeek, Kolibri-1 or MiniMax-M3 – or is the cloud API cheaper? With the local-vs-cloud AI inference calculator from ai-prices.eu you can compare hardware, electricity and maintenance costs against cloud API pricing and find your break-even point.
All Models in One Platform
With CompanyGPT you use all leading AI models GDPR-compliant in your own infrastructure:
- OpenAI GPT-6.1 Sol & GPT-6 Luna via Microsoft Foundry (EU Data Zone)
- Claude & Gemini via Google Vertex AI
- Claude & Llama via AWS Bedrock
- gpt-oss-120b & Llama via STACKIT
- Open models such as Kolibri-1 or Qwen self-hosted
- Full data control in your cloud
Request CompanyGPT
Use all leading AI models in one secure platform.
Discover CompanyGPT Request consultationWhich Model is Right for You?
We advise you on selecting the right AI models for your requirements.