Available AI Models

Delos gives you access to the market's best AI models from a single interface. Almost all models are served from European regions (France, Sweden, Norway, Switzerland, Northern Ireland), and a full range of open-source models is hosted in France at Scaleway.

Available models depend on your organization's plan. The "Plans" column in each table shows where each model is accessible:

Code	Plan
B	Base
Ω	Omega
O	Omicron
U	Upsilon
OS	Open Source

Not sure which plan you're on? Your administrator can see it in the organization settings.

Large Language Models (LLM)

Anthropic — Claude

Model	Plans	Hosting	Description
`claude-5-opus`	B Ω O U	🇪🇺 EU	Anthropic's most powerful model. Exceptional reasoning, complex analysis, and long-running tasks.
`claude-5-sonnet`	B Ω O U	🇪🇺 EU	Latest Sonnet generation. Excellent balance of performance and speed.
`claude-5.1-fable`	Ω	🇸🇪 Sweden	Fable model, geared toward writing and style.
`claude-5-fable`	Ω O U	🇸🇪 Sweden	Fable model, geared toward writing and content creation.
`claude-4.8-opus`	B Ω O U	🇪🇺 EU	Opus 4.8. Deep reasoning for complex tasks.
`claude-4.7-opus`	B Ω O U	🇸🇪 Sweden	Opus 4.7. Long-form analysis and in-depth work.
`claude-4.6-opus`	B Ω O U	🇫🇷 France	Opus 4.6. In-depth thinking and long-duration tasks.
`claude-4.6-sonnet`	B Ω O U OS	🇪🇺 EU	Sonnet 4.6. Fast and precise, ideal for code and analysis.
`claude-4.5-opus`	B Ω O U	🇫🇷 France	Opus 4.5. Tasks requiring deep thinking.
`claude-4.5-sonnet`	B Ω O U OS	🇫🇷 France	Sonnet 4.5. Versatile, excellent at writing and coding.
`claude-4.5-haiku`	B Ω O U OS	🇸🇪 Sweden	Anthropic's fastest and most economical model. Near-instant responses.

OpenAI — GPT

Model	Plans	Hosting	Description
`gpt-6-astra`	Ω	🇫🇷 France	OpenAI's latest generation. Advanced reasoning and top-tier creativity.
`gpt-5.6-terra`	B Ω O U	🇪🇺 EU	GPT-5.6 Terra. High-end general-purpose model.
`gpt-5.6-sol`	B Ω O U	🇪🇺 EU	GPT-5.6 Sol. Balance of reasoning and speed.
`gpt-5.6-luna`	B Ω O U OS	🇪🇺 EU	GPT-5.6 Luna. Fast, suited to everyday use.
`gpt-5.5`	B Ω O U	🇪🇺 EU	GPT-5.5. Stronger reasoning and better instruction-following.
`gpt-5.4-pro`	Ω	🇫🇷 France	Pro version of GPT-5.4. Maximum reasoning for hard problems.
`gpt-5.4`	B Ω O U	🇪🇺 EU	GPT-5.4. Complex analysis and demanding tasks.
`gpt-5.4-mini`	B Ω O U OS	🇪🇺 EU	Lighter version of GPT-5.4. Great value for money.
`gpt-5.4-nano`	Ω	🇫🇷 France	Ultra-lightweight. For simple tasks at very low latency.
`gpt-5.2`	B Ω O U OS	🇫🇷 France	GPT-5.2. Versatile and robust.
`gpt-5.1`	B Ω O U	🇪🇺 EU	GPT-5.1. Improved reasoning performance.
`gpt-5.0`	B Ω O U OS	🇪🇺 EU	GPT-5. Advanced reasoning and creativity.
`gpt-5.0-mini`	B Ω O U OS	🇪🇺 EU	Compact version of GPT-5.
`gpt-5.0-nano`	B Ω O U OS	🇪🇺 EU	Ultra-lightweight version of GPT-5. Fast and economical.
`gpt-4.1`	B Ω O U OS	🇪🇺 EU	Versatile, with a large context window. Ideal for long documents.
`gpt-4.1-mini`	B Ω O U OS	🇪🇺 EU	Lighter GPT-4.1. Speed and efficiency at a lower cost.
`gpt-4.1-nano`	B Ω O U OS	🇪🇺 EU	The fastest model in the GPT-4.1 family.
`gpt-4o`	B Ω O U OS	🇪🇺 EU	Omnimodal model. Text, images, and audio in a single conversation.
`gpt-4o-mini`	B Ω O U OS	🇫🇷 France	Compact GPT-4o. Fast and economical, multimodal capabilities.
`o3`	B Ω O U OS	🇪🇺 EU	Reasoning model. Excels at math, logic, and code.
`o4-mini`	B Ω O U OS	🇫🇷 France	Compact reasoning. Good depth-to-cost trade-off.

Google — Gemini

Model	Plans	Hosting	Description
`gemini-3.8-flash`	B Ω U	Google	Latest Flash. Very fast, massive context window.
`gemini-3.7-flash`	B Ω U	Google	Flash 3.7. Fast and multimodal.
`gemini-3.6-flash`	B Ω U	Google	Flash 3.6. Versatile and economical.
`gemini-3.5-flash`	B Ω U	Google	Flash 3.5. Good balance of speed and quality.
`gemini-3.5-flash-lite`	B Ω O U	Google	Lite version. Minimal latency for simple tasks.
`gemini-3.1-pro-preview`	B Ω U	Google	Gemini 3.1 Pro. Advanced reasoning and multimodality.
`gemini-3.1-flash-lite`	B Ω O U	Google	Flash Lite 3.1. Fast and economical.
`gemini-3.0-pro`	B Ω U	Google	Gemini 3 Pro. Massive context and multimodal capabilities.
`gemini-3.0-flash-preview`	B Ω U	Google	Gemini 3 Flash. Fast on large volumes.
`gemini-2.5-flash`	B Ω O U	Google	Gemini 2.5 Flash. Reliable for everyday tasks.

xAI — Grok

Model	Plans	Hosting	Description
`grok-4.6`	Ω	🇺🇸 / xAI	Latest Grok. Powerful, with real-time data access.
`grok-4.5`	Ω	xAI	Grok 4.5. Solid reasoning and fast responses.
`grok-4.3`	Ω	🇺🇸 United States	Grok 4.3. Strong general-purpose model.
`grok-4.1-fast`	Ω	xAI	Optimized for speed.
`grok-4.0-fast`	Ω	xAI	Grok 4, fast and economical.
`grok-build-0.1`	Ω	xAI	Model geared toward code generation.

Mistral

Model	Plans	Hosting	Description
`mistral-large-3`	B Ω O U OS	🇫🇷 France	Mistral AI's flagship model. Powerful and sovereign.
`mistral-medium-3-2505`	B Ω O U OS	🇫🇷 France	Mistral Medium. Good balance of performance and cost.
`mistral-small-3.2-2506`	B Ω O U OS	🇫🇷 Scaleway (fr-par)	Mistral Small 3.2. French sovereign and efficient model.

DeepSeek

Model	Plans	Hosting	Description
`deepseek-v4-pro`	Ω O U	Europe	DeepSeek V4 Pro. Excellent at reasoning and math.
`deepseek-v4.1-flash`	Ω O U	Europe	V4.1 Flash. Fast and economical.
`deepseek-v4-flash`	Ω O U	Europe	V4 Flash. Fast responses at low cost.
`deepseek-3.2`	B Ω O U OS	🇫🇷 France	DeepSeek 3.2, hosted in France.

Other models

Model	Provider	Plans	Hosting	Description
`kimi-k3.0`	Moonshot	Ω O U	Europe	Kimi K3. Very long context and solid reasoning.
`kimi-k2.6`	Moonshot	Ω O U	🇫🇷 France	Kimi K2.6. Versatile, good at document analysis.
`kimi-k2.5`	Moonshot	Ω O U	🇫🇷 France	Kimi K2.5. Efficient general-purpose model.
`glm-5.3`	Z.ai	Ω O U	Europe	GLM 5.3. Strong general-purpose model.
`glm-5.2`	Z.ai	Ω O U OS	🇫🇷 Scaleway (fr-par)	GLM 5.2, available with sovereign hosting.
`glm-4.7`	Z.ai	Ω O U	Europe	GLM 4.7. Good value for money.
`qwen3.8-max`	Alibaba	Ω O U	Europe	Qwen 3.8 Max. High-end model in the Qwen family.
`qwen3.6-plus`	Alibaba	Ω O U	Europe	Qwen 3.6 Plus. Multilingual general-purpose model.
`qwen3.0-235b`	Alibaba	B Ω O U OS	🇫🇷 Scaleway (fr-par)	Qwen 3, 235B parameters, MoE architecture, EU-sovereign.
`qwen3.0-32b`	Alibaba	B Ω O U OS	🇸🇪 Sweden	Qwen 3 32B. Compact and capable.
`llama-4-maverick`	Meta	B Ω O U OS	🇫🇷 France	Llama 4 Maverick. MoE architecture, high performance.
`llama-4-scout`	Meta	B Ω O U OS	🇫🇷 France	Llama 4 Scout. Efficient open-source model.
`nova-pro`	Amazon	B Ω O U OS	🇫🇷 France	Amazon Nova Pro. Multimodal and versatile.
`nova-lite`	Amazon	B Ω O U OS	🇫🇷 France	Nova Lite. Fast and economical.
`nova-2-lite`	Amazon	B Ω O U OS	🇫🇷 France	Nova 2 Lite. Latest lightweight generation.

Sovereign open-source models (Scaleway 🇫🇷)

These models run on Scaleway servers in Paris (fr-par). They are included in the Open Source plan.

Model	Provider	Plans	Description
`qwen3.5-397b-a17b`	Alibaba	OS	Qwen 3.5, 397B MoE parameters. The most capable in the sovereign range.
`qwen3.6-35b-a3b`	Alibaba	OS	Compact Qwen 3.6 MoE. Fast and efficient.
`qwen3-235b-a22b-instruct-2507`	Alibaba	OS	Qwen 3 235B MoE, EU-sovereign.
`qwen3-coder-30b-a3b-instruct`	Alibaba	OS	Specialized for code generation.
`devstral-2-123b-instruct-2512`	Mistral	OS	Devstral 2 123B. Geared toward software development.
`devstral-2-2512`	Mistral	OS	Devstral 2. Sovereign coding assistant.
`mistral-small-3.2-24b-instruct-2506`	Mistral	OS	Mistral Small 3.2 24B. French and sovereign.
`pixtral-12b-2409`	Mistral	OS	Pixtral 12B. Vision-language model.
`llama-3.3-70b-instruct`	Meta	Ω OS	Llama 3.3 70B. High-performing open-source model.
`llama-3.1-8b-instruct`	Meta	OS	Llama 3.1 8B. Compact, fast, and economical.
`gemma-4-26b-a4b-it`	Google	OS	Gemma 4 26B MoE. Lightweight and capable.
`gemma-3-27b-it`	Google	OS	Gemma 3 27B. Solid open-source generalist.
`gpt-oss-120b`	OpenAI	Ω O U OS	Open-source 120B model. Sovereign alternative.

Image generation

Model	Provider	Plans	Hosting	Description
`gpt-image-2`	OpenAI	B Ω O U OS	🇸🇪 Sweden	Latest OpenAI image generation. Realism and prompt fidelity.
`gpt-image-1`	OpenAI	B Ω O U OS	🇸🇪 Sweden	OpenAI image generation. Creative and realistic visuals.
`gemini-3.0-pro-image`	Google	B Ω U	Google	Advanced image generation and editing by Gemini.
`gemini-3.1-flash-image`	Google	B Ω U	Google	Fast image generation.
`gemini-2.5-flash-image`	Google	B Ω U	Google	Gemini 2.5 image generation.
`grok-imagine-image-2.0`	xAI	Ω	xAI	Image generation by xAI.

Video generation

Model	Provider	Plans	Hosting	Description
`veo-3.1-generate-preview`	Google	B Ω U	Google	Veo 3.1. High-quality text-to-video generation.
`veo-3.1-fast-generate-preview`	Google	B Ω U	Google	Veo 3.1 Fast. Accelerated video generation.
`veo-3.1-lite-generate-preview`	Google	B Ω U	Google	Veo 3.1 Lite. Lighter and economical version.
`grok-imagine-video-1.5`	xAI	Ω	xAI	Video generation by xAI.
`gemini-omni-1.1-flash`	Google	B Ω U	Google	Gemini omnimodal model, including video.

Audio transcription

Model	Provider	Plans	Hosting	Description
`gpt-4o-transcribe`	OpenAI	B Ω O U OS	🇫🇷 France	Advanced transcription based on GPT-4o, with contextual understanding.
`whisper-large-v3`	OpenAI	B Ω O U OS	🇫🇷 Scaleway (fr-par)	Sovereign Whisper Large V3. Accurate, multilingual transcription.
`whisper-1`	OpenAI	B Ω O U OS	🇸🇪 Sweden	Automatic speech transcription. Dozens of languages.
`voxtral-mini-transcribe`	Mistral	B Ω O U OS	Mistral	Voxtral transcription. Advanced audio understanding.

Text-to-speech (TTS)

Model	Provider	Plans	Hosting	Description
`eleven_v3`	ElevenLabs	Ω U	ElevenLabs	Latest ElevenLabs generation. Very natural, expressive voices.
`eleven_multilingual_v2`	ElevenLabs	Ω U	ElevenLabs	High-quality multilingual text-to-speech.
`eleven_turbo_v2_5`	ElevenLabs	Ω U	ElevenLabs	Turbo version. Low latency for real-time use.
`gemini-3.1-flash-tts-preview`	Google	B Ω U	Google	Gemini 3.1 text-to-speech.
`gemini-2.5-pro-preview-tts`	Google	B Ω U	Google	Gemini 2.5 Pro text-to-speech.
`gemini-2.5-flash-preview-tts`	Google	B Ω U	Google	Fast text-to-speech.
`grok-tts`	xAI	Ω	xAI	Text-to-speech by xAI.

Advanced audio processing

Model	Provider	Plans	Hosting	Description
`eleven_dubbing`	ElevenLabs	Ω U	ElevenLabs	Automatic video dubbing into another language.
`eleven_audio_isolation`	ElevenLabs	Ω U	ElevenLabs	Voice isolation and background noise removal.

Embeddings

Model	Provider	Plans	Hosting	Description
`text-embedding-3-large`	OpenAI	B Ω O U OS	🇪🇺 EU	High-dimensional embeddings for semantic search and RAG.
`text-embedding-3-small`	OpenAI	B Ω O U OS	🇪🇺 EU	Compact embeddings. Fast and efficient.
`qwen3-embedding-8b`	Alibaba	OS	🇫🇷 Scaleway (fr-par)	Sovereign embeddings, hosted in France.
`bge-multilingual-gemma2`	Google	OS	🇫🇷 Scaleway (fr-par)	Sovereign multilingual embeddings. Ideal for semantic search.

Vision

Model	Provider	Plans	Hosting	Description
`azure_vision`	Microsoft	B Ω O U OS	🇫🇷 France	Image analysis and OCR. Text and structure extraction.

Notes

Hosting. Anthropic, OpenAI, Mistral, Meta, Amazon models and the entire open-source range are served from European regions. Google models (Gemini, Veo) and xAI (Grok) are served by their respective providers.
Sovereignty. Models marked "Scaleway (fr-par)" run on French servers. This is the range to favor for use cases with sovereignty requirements.
High availability. Several models have redundant deployments across multiple regions: in the event of an incident, requests fail over automatically, with no action needed on your part.