Skip to main content

Catalog of LLM as a Service Models

Overview​

Cloud Temple LLMaaS offers 56 large language models carefully selected and optimized to meet the strictest SecNumCloud requirements. Our catalog covers the entire spectrum, from ultra-efficient micro-models to extremely large models.

Global Statistics​

MetricValue
Total number of models56 models
Minimum context512 tokens
Maximum context1,000,000 tokens
ComplianceSecNumCloud ✅ HDS ✅ Sovereignty ✅ C5 ✅
Location100% France 🇫🇷

Pricing​

Usage TypePrice
Input tokens1.8€ / million tokens
Output tokens8€ / million tokens
Advanced reasoning8€ / million tokens
Document reranking4€ / million reranked tokens
Asynchronous batch (input)0.9€ / million tokens
Asynchronous batch (output)4€ / million tokens

Chat & Reasoning Models​

General-purpose models for dialogue, analysis, reasoning, and multilingual tasks.

cogito:32b​

Deep Cogito • 32B parameters • Context : 32,000 tokens

Advanced analytical reasoning model, designed for decomposing complex problems and logical verification.

Technical specifications:

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Safety

Tags : Agent Raisonnement Compréhension Analyse

Use cases :

  • Multi-factorial scenario analysis
  • Scientific problem solving with formal demonstration
  • Expert systems (juridique, médical, technique)

gemma3:27b​

Google • 27B parameters • Context : 120,000 tokens

Google's multimodal model with integrated vision and support for 140+ languages. 120K token context.

Technical specifications:

  • Speed : 23 tokens/second
  • Consumption : 5.80 kWh/million tokens
  • License : Google Gemma Terms of Use
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Safety

Tags : Vision Agent Long Context

Use cases :

  • Document analysis with extended context (120K tokens)
  • Simultaneous processing of images and text
  • Structured extraction from PDFs and scanned documents

glm-4.7-flash:30b​

Zhipu AI • 30B parameters • Context : 120,000 tokens

Fast model with an excellent performance/latency balance for reasoning and analysis.

Technical specifications:

  • Speed : 88 tokens/second
  • Consumption : 1.58 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Security

Tags : Agent Fast Large Context Multilingual

Use cases :

  • Fast conversational assistants
  • Long document analysis (120k tokens)
  • Reasoning tasks with low latency

gpt-oss:120b​

OpenAI • 120B parameters • Context: 120,000 tokens

State-of-the-art open-weight model from OpenAI with configurable reasoning and transparent chain of thought.

Technical specifications:

  • Speed : 94 tokens/second
  • Consumption : 2.37 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Safety

Tags: MoE Agent Reasoning Open-Source Very Large

Use cases:

  • Advanced conversational agents with complex reasoning
  • Applications requiring transparency in the reasoning process
  • Business scenarios requiring a permissive license

gpt-oss:20b​

OpenAI • 20B parameters • Context: 120,000 tokens

Compact version of the OpenAI model, optimized for fast inference with strong reasoning capabilities.

Technical specifications:

  • Speed : 41 tokens/second
  • Consumption : 3.25 kWh/million tokens
  • License : Apache 2.0
  • Region : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Security

Tags : MoE Agent Reasoning Open-Source Compact Fast

Use cases :

  • Fast inference with strong reasoning capabilities
  • Agentic use cases (function calling, web navigation, code execution)
  • Applications with limited compute budget

llama3.3:70b​

Meta • 70B parameters • Context: 132,000 tokens

Meta's multilingual model, excelling in natural dialogue and nuanced understanding across 8 languages.

Technical specifications:

Capabilities: ✅ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Safety

Tags: Agent Dialogue Multilingual

Use cases:

  • Multilingual chatbots (8 langues)
  • Analysis of large legal or technical documents
  • Generation of structured texts with high stylistic fidelity

ministral-3:14b​

Mistral AI • 14B parameters • Context: 250,000 tokens

The most powerful in the Ministral family, with advanced reasoning and coding. 250K token context.

Technical specifications:

  • Speed : 28 tokens/second
  • Consumption : 4.74 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Security

Tags: High Performance Reasoning Code

Use cases:

  • Complex problem solving
  • Coding and engineering assistants
  • In-depth document analysis with reasoning

ministral-3:3b​

Mistral AI • 3B parameters • Context : 250,000 tokens

Compact Mistral model, high-performing despite its small size. 250K token context.

Technical specifications:

  • Speed : 22 tokens/second
  • Consumption : 1.75 kWh/million tokens
  • License : Apache 2.0
  • Location : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Compact Efficient

Use cases :

  • Reactive personal assistants
  • Fast classification and routing
  • Simple tasks with large context

ministral-3:8b​

Mistral AI • 8B parameters • Context: 250,000 tokens

Intermediate Ministral model with an excellent performance/speed trade-off. 250K token context.

Technical specifications:

  • Speed : 40 tokens/second
  • Consumption : 3.33 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Safety

Tags: Efficient Reasoning

Use cases:

  • Advanced conversational assistants
  • Document analysis and information extraction
  • Good speed/quality trade-off

mistral-small3.2:24b​

Mistral AI • 24B parameters • Context: 128,000 tokens

Mistral model with enhanced instruction following, robust function calling, and vision capabilities. Integrated problematic content detection.

Technical specifications:

  • Speed : 28 tokens/second
  • Consumption : 5.05 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ✅ Security

Tags : Vision Agent Security Instruction Following

Use cases :

  • Reliable conversational agents with precise instruction following
  • External tool integration via function calling
  • Applications requiring native security filtering

mistral-small4:119b​

Mistral AI • 119B parameters • Context : 262,144 tokens

High-performance Mistral model (119B) with vision, built-in security, and a 262K token context. Fast (100 t/s).

Technical specifications:

  • Speed : 100 tokens/second
  • Consumption : 2.00 kWh/million tokens
  • License : Apache 2.0
  • Location : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ✅ Security

Tags: Vision Agent Security Large Context Fast

Use cases:

  • High-performance conversational agents with vision
  • Analysis of very long documents (262K tokens)
  • Critical applications requiring reliability and security

nemotron-3-super:120b​

NVIDIA • 120B parameters • Context: 1,000,000 tokens

NVIDIA model optimized for collaborative agents, long reasoning, and high-volume workloads. 1M token context.

Technical specifications:

  • Speed : 72 tokens/second
  • Consumption : 1.93 kWh/million tokens
  • License : NVIDIA Community License
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Security

Tags : Agent Reasoning Long Context

Use cases :

  • Autonomous agents with multiple tool calls
  • High-volume workflow automation
  • Long document analysis with precise extraction

nemotron-cascade:30b​

NVIDIA • 30B parameters • Context: 1,000,000 tokens

NVIDIA model specialized in mathematics (IMO 2025 gold medal) and problem decomposition. 1M token context.

Technical specifications:

  • Speed: 130 tokens/second
  • Consumption: 1.93 kWh/million tokens
  • License: NVIDIA Community License
  • Localization: FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Safety

Tags: Agent Reasoning Long Context Math

Use cases:

  • Solving complex mathematical problems
  • Agents with multiple tool calls
  • Long document analysis

nemotron3-nano:30b​

NVIDIA • 30B parameters • Context: 1,000,000 tokens

Ultra-fast NVIDIA model (160 t/s) with reasoning and function calling. 1M token context.

Technical Specifications:

  • Speed : 160 tokens/second
  • Consumption : 1.56 kWh/million tokens
  • License : NVIDIA Community License
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Safety

Tags: Agent Reasoning Long Context Fast

Use Cases :

  • Autonomous agents requiring fast responses
  • Logical reasoning and problem solving
  • Long document analysis with precise extraction

olmo-3:32b​

AllenAI • 32B parameters • Context : 65,536 tokens

First fully open reasoning model at this scale. Total transparency (data, code, weights).

Technical specifications:

  • Speed : 22 tokens/second
  • Consumption : 5.98 kWh/million tokens
  • License : Apache 2.0
  • Location : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Security

Tags : Open-Source Large Context Reasoning Transparent Code High Performance

Use cases :

  • Complex reasoning and multi-step problem solving
  • Software development with decision process transparency
  • Critical tasks requiring full auditability

olmo-3:7b​

AllenAI • 7B parameters • Context : 65,536 tokens

Fully open and efficient model, excelling in mathematics and programming with complete transparency.

Technical specifications:

  • Speed : 35 tokens/second
  • Consumption : 1.13 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Open-Source Grand Contexte Transparent Efficient Maths Code

Use cases :

  • Academic research requiring full reproducibility
  • Programming and mathematical problem solving
  • Document analysis with complete traceability

qwen3-2507-think:4b​

Qwen Team • 4B parameters • Context: 250,000 tokens

Compact model optimized for deep reasoning (logic, math, science, code). 250K token context.

Technical specifications:

  • Speed : 55 tokens/second
  • Consumption : 2.42 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Safety

Tags : Agent Reasoning Large Context Compact Fast

Use cases :

  • Complex reasoning (logic, math, science, code)
  • Agents with long conversation history (250K tokens)
  • Analysis of large documents with deep reasoning

qwen3-2507:235b​

Qwen Team • 235B parameters • Context: 200,000 tokens

The most powerful model in the catalog (235B parameters, 22B active). Excels in mathematics, coding, and logical reasoning.

Technical specifications:

  • Speed: 56 tokens/second
  • Consumption: 3.97 kWh/million tokens
  • License: Apache 2.0
  • Location: FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Safety

Tags: MoE Agent Reasoning Very Large

Use cases:

  • Solving complex mathematical and logical problems
  • Tasks requiring a vast knowledge base
  • Advanced coding assistant

qwen3-omni:30b​

Qwen Team • 30B parameters • Context: 32,768 tokens

Native omni-modal model — simultaneously understands text, images, video, and audio.

Technical specifications:

  • Speed: 19 tokens/second
  • Consumption: 7.43 kWh/million tokens
  • License: Apache 2.0
  • Localization: FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ✅ Reasoning • ❌ Safety

Tags: Omni Audio Vision Agent Multimodal

Use cases:

  • Multimodal interactions (text + image + audio + video)
  • Combined video and audio analysis
  • Next-generation intelligent assistants

qwen3.5:0.8b​

Qwen Team • 0.8B parameters • Context : 250,000 tokens

Ultra-lightweight model with an exceptional 250K token context — remarkable for a model of this size.

Technical specifications:

  • Speed : 16 tokens/second
  • Consumption : 2.39 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Compact Efficient Grand Contexte Multilingue

Use cases :

  • Text processing with very large context (250K tokens)
  • Fast classification and sorting
  • Lightweight assistants with large history

qwen3.5:4b​

Qwen Team • 4B parameters • Context : 250,000 tokens

Compact Qwen3.5 model with a good performance/efficiency balance and a 250K token context.

Technical specifications:

  • Speed : 37 tokens/second
  • Consumption : 3.64 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Safety

Tags : Compact Efficient Large Context Multilingual

Use cases :

  • Conversational assistants with large context
  • Text summarization and classification
  • Simple coding tasks

qwen3.5:9b​

Qwen Team • 9B parameters • Context: 250,000 tokens

Intermediate Qwen3.5 model with solid reasoning and extended context up to 250K tokens.

Technical specifications:

  • Speed: 32 tokens/second
  • Consumption: 4.23 kWh/million tokens
  • License: Apache 2.0
  • Localization: FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Safety

Tags: Efficient Long Context Multilingual Reasoning

Use cases:

  • Advanced conversational assistants
  • Document analysis and information extraction
  • Medium-complexity reasoning tasks

qwen3.6:27b​

Qwen Team • 27B parameters • Context : 1,000,000 tokens

General-purpose reference model with a native 1M token context. Excels in reasoning, instruction following, and multilingual tasks.

Technical Specifications:

  • Speed : 80 tokens/second
  • Consumption : 2.78 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ✅ Reasoning • ❌ Security

Tags : Agent Large Context Multilingual Vision Reasoning

Use Cases :

  • Versatile assistants with precise instruction following
  • Analysis of very large documents (1M tokens)
  • Multilingual applications with logical reasoning
  • Programming and code generation

qwen3:0.6b​

Qwen Team • 0.6B parameters • Context: 40,000 tokens

Ultra-fast micro-model for simple tasks and routing. 40K token context.

Technical specifications:

  • Speed : 46 tokens/second
  • Consumption : 1.33 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Compact Fast Efficient Multilingual

Use cases :

  • Fast classification and sorting
  • Lightweight assistants with low latency
  • Routing in multi-model architectures

Programming & Agent Models​

Models specialized in code, software engineering, and autonomous agents.

devstral-small-2:24b​

Mistral AI & All Hands AI • 24B parameters • Context : 200,000 tokens

State-of-the-art agentic model for software engineering. Near >100B model performance for code. Integrated vision.

Technical specifications:

  • Speed : 33 tokens/second
  • Consumption : 4.23 kWh/million tokens
  • License : Apache 2.0
  • Location : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ✅ Security

Tags : Agent Programming Vision Open-Source Long Context Fast

Use cases :

  • Autonomous coding agents
  • Rapid code refactoring
  • Iterative engineering tasks

functiongemma:270m​

Google • 270M parameters • Context: 32,768 tokens

Micro-model specialized in function call detection. Ideal as a router in an agentic architecture.

Technical specifications:

  • Speed : 40 tokens/second
  • Consumption : 0.97 kWh/million tokens
  • License : Google Gemma Terms of Use
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags: Agent Compact Efficient Function Calling

Use cases:

  • Tool router in an agentic architecture
  • Function call intent detection
  • Fast pre-filtering before routing

qwen-coder-next:80b​

Qwen Team • 80B parameters • Context : 250,000 tokens

State-of-the-art model for code and complex reasoning. 250K token context.

Technical specifications:

  • Speed : 97 tokens/second
  • Consumption : 2.29 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Safety

Tags : Agent Programming MoE Long Context

Use cases :

  • Advanced programming assistant (repository scale)
  • Complex code analysis and refactoring
  • Autonomous software engineering agents

qwen3-next:80b​

Qwen Team • 80B parameters • Context: 250,000 tokens

Versatile 80B model optimized for long contexts, function calling, and structured reasoning.

Technical specifications:

  • Speed : 67 tokens/second
  • Consumption : 2.09 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags: Agent Grand Contexte MoE

Use cases:

  • Advanced conversational agents with tool integration
  • Analysis of very large documents
  • Code generation and structured reasoning

qwen3.6:35b​

Qwen Team • 35B parameters • Context: 1,000,000 tokens

Leader in agentic software engineering (SWE-bench 73.4%). 1M token context, integrated vision and tool calling.

Technical specifications:

  • Speed : 121 tokens/second
  • Consumption : 2.07 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ✅ Reasoning • ❌ Security

Tags : Agent Programming Long Context MoE Vision Reasoning

Use cases :

  • Agentic coding assistant in IDEs (Cursor, Continue.dev, VS Code Copilot)
  • Analysis of complete codebases (1M tokens)
  • Automated code review and intelligent CI/CD
  • Complex bug resolution with multi-step reasoning

rnj-1:8b​

Essential AI • 8B parameters • Context : 32,000 tokens

STEM-specialized model — excels in code (83.5% HumanEval+), mathematics, and sciences.

Technical specifications:

  • Speed : 23 tokens/second
  • Consumption : 1.69 kWh/million tokens
  • License : Open Weights
  • Location : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ✅ Reasoning • ❌ Security

Tags: Code Maths STEM Reasoning Efficient

Use cases:

  • Advanced code generation
  • Solving complex mathematical problems
  • Scientific and technical tasks

Vision & Multimodal Models​

Models for image, video, OCR, and visual document analysis.

deepseek-ocr​

DeepSeek AI • 3B parameters • Context: 8,192 tokens

Specialized OCR model for high-precision text extraction with layout preservation (tables, formulas).

Technical specifications:

  • Speed : 84 tokens/second
  • Consumption : 0.66 kWh/million tokens
  • License : MIT license
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Security

Tags : Vision OCR Efficient

Use cases :

  • Structured text extraction (Markdown/LaTeX) from images/PDFs
  • Document scanning with tables and formulas

gemma4:31b​

Google • 31B parameters • Context: 250,000 tokens

Google's dense multimodal model, ranked 3rd globally on Arena AI. Advanced vision, reasoning, and coding. 250K token context.

Technical Specifications:

  • Speed: 59 tokens/second
  • Consumption: 3.77 kWh/million tokens
  • License: Apache 2.0
  • Location: FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ✅ Reasoning • ❌ Safety

Tags: Agent Grand Contexte Vision Raisonnement Multilingue Open-Source

Use Cases:

  • Autonomous agents with reasoning and tool calling
  • Analysis of large documents with vision
  • Visual understanding (OCR, graphiques, documents scannés)

gemma4:e2b​

Google • 31B (E2B) parameters • Context: 128,000 tokens

Ultra-fast (125 t/s) variant of Gemma 4 with vision. Excellent energy efficiency.

Technical specifications:

  • Speed: 125 tokens/second
  • Consumption: 1.11 kWh/million tokens
  • License: Apache 2.0
  • Location: FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Safety

Tags: Vision Fast Efficient

Use cases:

  • Low-latency image analysis
  • Large documents with reduced footprint

gemma4:e4b​

Google • 31B (E4B) parameters • Context : 128,000 tokens

Variant of Gemma 4 with a better quality/speed trade-off than the E2B version. Integrated vision.

Technical specifications :

  • Speed : 85 tokens/second
  • Consumption : 1.63 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities : ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Safety

Tags : Vision Fast

Use cases :

  • Document and image analysis with a good quality/speed trade-off
  • Fast alternative for vision tasks

granite3.2-vision:2b​

IBM • 2B parameters • Context: 16,384 tokens

Compact IBM Granite vision model for fast OCR and data extraction from scanned documents.

Technical specifications:

  • Speed : 49 tokens/second
  • Consumption : 0.80 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Security

Tags: Vision Compact Efficient OCR

Use cases:

  • Fast OCR on lightweight documents
  • Data extraction from images
  • Low-consumption visual analysis

qwen3-vl:235b​

Qwen Team • 235B parameters • Context : 200,000 tokens

The most powerful multimodal model in the catalog. State-of-the-art visual understanding and exceptional reasoning.

Technical specifications:

  • Speed : 24 tokens/second
  • Consumption : 5.56 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ✅ Reasoning • ❌ Security

Tags : Agent Reasoning Long Context Vision

Use cases :

  • Complex document automation (multilingual OCR, structured extraction)
  • Intelligent visual agents
  • Advanced scientific analysis (STEM, spatial reasoning)
  • Multimodal RAG on documents and videos

qwen3-vl:2b​

Qwen Team • 2B parameters • Context: 250,000 tokens

Ultra-compact vision model for fast OCR, object detection, and embedded applications.

Technical specifications:

  • Speed : 64 tokens/second
  • Consumption : 0.95 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Security

Tags: Vision Compact Efficient Multimodal OCR

Use cases:

  • Real-time image analysis
  • Lightweight document OCR and reading
  • Fast visual sorting and classification

qwen3-vl:30b​

Qwen Team • 30B parameters • Context: 250,000 tokens

High-performance multimodal model for OCR, object detection, video analysis, and spatio-temporal reasoning.

Technical specifications:

  • Speed : 39 tokens/second
  • Consumption : 3.39 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Safety

Tags : Vision Agent Long Context Multimodal Video OCR

Use cases :

  • Long video analysis and intelligent surveillance
  • Structured data extraction (documents, tables, charts)
  • Visual assistants with spatial understanding

qwen3-vl:32b​

Qwen Team • 32B parameters • Context : 250,000 tokens

High-performance variant for the most demanding vision tasks. 250K token context.

Technical specifications:

  • Speed : 17 tokens/second
  • Consumption : 7.75 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Safety

Tags : Vision Agent Long Context Multimodal Video OCR

Use cases :

  • Scientific and technical analysis of high-resolution images
  • Automation of complex visual processes
  • Detailed understanding of dynamic scenes

qwen3-vl:4b​

Qwen Team • 4B parameters • Context : 250,000 tokens

Compact and fast vision model for document analysis and video understanding.

Technical specifications:

  • Speed : 57 tokens/second
  • Consumption : 2.34 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Safety

Tags : Vision Compact Multimodal Efficient Video OCR

Use cases :

  • Automated document analysis (invoices, forms)
  • Video content understanding
  • Interactive visual assistants

qwen3-vl:8b​

Qwen Team • 8B parameters • Context: 250,000 tokens

Intermediate vision model — a good balance between performance and footprint. 250K token context.

Technical specifications:

  • Speed : 39 tokens/second
  • Consumption : 3.38 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ✅ Tools/Agent • ✅ Vision • ❌ Reasoning • ❌ Security

Tags: Vision Multimodal Efficient Video OCR

Use cases:

  • Automated document analysis
  • Video content understanding
  • Interactive visual assistants

Embedding Models​

Models for semantic search and Retrieval-Augmented Generation (RAG).

bge-m3:567m​

BAAI • 567M parameters • Context : 8,192 tokens

State-of-the-art multilingual embedding (100+ languages). Supports dense, sparse, and multi-vector search.

Technical specifications:

  • Speed : 171 tokens/second
  • Consumption : 0.36 kWh/million tokens
  • License : MIT
  • Location : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags: Embedding Multilingual Efficient

Use cases:

  • Multilingual semantic search
  • Retrieval-Augmented Generation (RAG)
  • Document clustering and classification

embeddinggemma:300m​

Google • 300M parameters • Context: 2,048 tokens

Google multilingual embedding (100+ languages), optimized for search and semantic retrieval.

Technical specifications:

  • Speed : 175 tokens/second
  • Consumption : 0.35 kWh/million tokens
  • License : Google Gemma Terms of Use
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags: Embedding Compact Semantic Efficient Multilingual

Use cases:

  • Search and information retrieval
  • Document classification and clustering
  • Semantic similarity search

granite-embedding:278m​

IBM • 278M parameters • Context: 512 tokens

Ultra-compact IBM embedding for minimal-latency semantic search.

Technical specifications:

  • Speed : 196.3 tokens/second
  • Consumption : 0.31 kWh/million tokens
  • License : Apache 2.0
  • Location : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags: Embedding Compact Efficient

Use cases:

  • High-frequency semantic search
  • Document clustering

qwen3-embedding:0.6b​

Qwen Team • 0.6B parameters • Context : 32,768 tokens

Ultra-lightweight and fast embedding for low-latency semantic search.

Technical specifications:

  • Speed : N/A
  • Consumption : 0.57 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Embedding Compact Efficient

Use cases :

  • Fast semantic search
  • Real-time text classification

qwen3-embedding:4b​

Qwen Team • 4B parameters • Context: 40,000 tokens

High-performance embedding with deep semantic understanding and extended context (40K tokens).

Technical specifications:

  • Speed : N/A
  • Consumption : 0.57 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Embedding Large Context Efficient

Use cases :

  • Semantic search on long documents
  • RAG with extended context
  • High-precision semantic analysis

qwen3-embedding:8b​

Qwen Team • 8B parameters • Context: 40,000 tokens

High-capacity embedding with the best semantic understanding in the Qwen3 family. Extended context (40K tokens).

Technical specifications:

  • Speed : N/A
  • Consumption : 0.57 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Embedding Large Context High Performance

Use cases :

  • High-precision semantic search
  • Advanced RAG with extended context
  • Semantic analysis of complex documents

Reranking Models​

Models for reranking results in RAG pipelines.

bge-reranker-large​

BAAI • 335M parameters • Context: 512 tokens

High-performance reranking model from the BGE family, multilingual.

Technical specifications:

  • Speed : N/A
  • Consumption : N/A
  • License : MIT
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Reranker High Performance

Use Cases :

  • Multilingual reranking for RAG

nvidia/llama-nemotron-rerank-vl-1b-v2​

NVIDIA • 1B parameters • Context : 4,096 tokens

Cohere API-compatible reranking model (/v1/rerank and /v2/rerank). Ranks documents by relevance to a query.

Technical specifications:

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Safety

Tags: Rerank RAG Compact

Use cases:

  • Reordering of semantic search results
  • Refining relevance in RAG pipelines

qwen3-reranker:0.6b​

Qwen Team • 0.6B parameters • Context: 4,096 tokens

Compact and efficient reranking model for fast reordering.

Technical specifications:

  • Speed : N/A
  • Consumption : N/A
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Safety

Tags : Reranker Compact Efficient

Use cases :

  • Fast reordering for RAG

qwen3-reranker:4b​

Qwen Team • 4B parameters • Context : 4,096 tokens

High-performance reranking model with strong contextual understanding capabilities.

Technical specifications:

  • Speed : N/A
  • Consumption : N/A
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Reranker Performance

Use cases:

  • High-quality reranking for RAG

Security Models​

Models for content filtering, compliance, and guardrails.

granite3-guardian:2b​

IBM • 2B parameters • Context: 8,192 tokens

Compact version of the Granite Guardian security model for low-latency filtering.

Technical specifications:

  • Speed : 60 tokens/second
  • Consumption : 0.65 kWh/million tokens
  • License : Apache 2.0
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ✅ Security

Tags: Security Guardrails Compact Efficient

Use cases:

  • Low-latency guardrails in agentic workflows
  • Real-time security filtering

granite3-guardian:8b​

IBM • 8B parameters • Context: 8,192 tokens

Security model specialized in detecting problematic content, jailbreaks, and regulatory compliance.

Technical specifications:

  • Speed: 45 tokens/second
  • Consumption: 3.09 kWh/million tokens
  • License: Apache 2.0
  • Localization: FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ✅ Security

Tags: Security Guardrails Compliance Filtering

Use cases:

  • Detection of problematic content and jailbreaks
  • Pre/post-generation security filtering
  • Regulatory compliance (RGPD, HDS)

Translation Models​

Models specialized in high-fidelity multilingual translation.

translategemma:12b​

Google • 12B parameters • Context: 128,000 tokens

High-fidelity translation for 55 languages with a 128K token context.

Technical Specifications:

  • Speed: 27 tokens/second
  • Consumption: 4.87 kWh/million tokens
  • License: Gemma Terms of Use
  • Localization: FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags: Translation Multilingual Specialized

Use Cases:

  • Long document translation
  • Cross-language communication
  • Content localization

translategemma:27b​

Google • 27B parameters • Context : 120,000 tokens

High-performance translation for 55 languages. Superior quality for complex and technical content.

Technical specifications:

  • Speed : 17 tokens/second
  • Consumption : 7.84 kWh/million tokens
  • License : Gemma Terms of Use
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Safety

Tags: Translation Multilingual Specialized High Performance

Use cases:

  • High-precision translation
  • Technical document translation
  • Literary and cultural nuances

translategemma:4b​

Google • 4B parameters • Context : 128,000 tokens

Fast and efficient translation for 55 languages. Ideal for real-time localization.

Technical specifications:

  • Speed : 31 tokens/second
  • Consumption : 1.25 kWh/million tokens
  • License : Gemma Terms of Use
  • Localization : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags : Translation Multilingual Specialized Efficient

Use cases :

  • Fast text translation
  • Real-time localization
  • Budget-limited translation

Audio & Image Models​

Models for audio transcription and image generation.

voxtral​

Mistral AI • 4B parameters • Context: 32,768 tokens

Real-time audio transcription via WebSocket. Streaming speech recognition with low latency.

Technical specifications:

  • Speed: N/A
  • Consumption: N/A
  • License: Apache 2.0
  • Region: FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Security

Tags: ASR Audio Realtime WebSocket

Use cases:

  • Real-time audio transcription (streaming)
  • Voice assistants
  • Live captioning

z-image:16b​

Community • 16B parameters • Context : 0 tokens

Image generation from text prompts, compatible with OpenAI API /v1/images/generations.

Technical specifications:

  • Speed : N/A
  • Consumption : N/A
  • License : Open Weights
  • Location : FR 🇫🇷

Capabilities: ❌ Tools/Agent • ❌ Vision • ❌ Reasoning • ❌ Safety

Tags : Image Generation Creative Multimodal

Use cases :

  • Image generation from text descriptions
  • Visual content creation
  • Rapid visual prototyping

Multilingual Dialogue​

Chatbots and assistants capable of communicating in multiple languages with automatic detection and context preservation

Recommended models:

  • nemotron-3-super:120b
  • qwen3.6:27b
  • nemotron3-nano:30b
  • gpt-oss:120b

Long document analysis​

Processing large documents (>100 pages) with key information extraction, summaries, and question answering

Recommended models:

  • nemotron-3-super:120b
  • qwen3.6:27b
  • qwen3-2507:235b

Programming and development​

Code generation, optimization, and debugging across multiple languages, refactoring, and test creation

Recommended models:

  • qwen3.6:35b
  • qwen-coder-next:80b
  • devstral-small-2:24b
  • nemotron-3-super:120b

Visual Analysis​

Image and visual document processing, OCR, chart and table interpretation

Recommended models:

  • qwen3-vl:235b
  • gemma4:31b
  • deepseek-ocr
  • qwen3-vl:30b

Security and Compliance​

Sensitive content filtering, jailbreak detection, GDPR/HDS compliance

Recommended models:

  • granite3-guardian:8b
  • granite3-guardian:2b
  • mistral-small4:119b

Lightweight Deployments​

Applications requiring minimal footprint, low latency, and low resource consumption

Recommended models:

  • qwen3.5:0.8b
  • qwen3-vl:2b
  • ministral-3:3b

RAG (Retrieval-Augmented Generation)​

Complete pipelines for semantic search, reranking, and retrieval-augmented generation

Recommended models:

  • bge-m3:567m
  • nvidia/llama-nemotron-rerank-vl-1b-v2
  • qwen3.6:27b