Our Large Language Model as a Service (LLMaaS) offering gives you access to cutting-edge language models, inferred using SecNumCloud-qualified infrastructure, HDS-certified for healthcare data hosting, and therefore sovereign, calculated in France. Benefit from high performance and optimal security for your AI applications. Your data remains strictly confidential, and is neither exploited nor stored after processing.
Simple, transparent pricing
1.8
€
per million input tokens
8
€
per million tokens issued
8
€
per million reasoning tokens
0,01 €
per minute of transcribed audio *
Calculated on an infrastructure based in France, SecNumcloud qualified and HDS certified.
Note on the "Reasoning" price:
This price applies specifically to models classified as "reasoners" or "hybrids" (models with the "Reasoning" capability activated) when reasoning is active and only on tokens linked to this activity.
* any minute started is counted
Large models
Our large models offer state-of-the-art performance for the most demanding tasks. They are particularly well-suited to applications requiring a deep understanding of language, complex reasoning or the processing of long documents.
Specialised models
Our specialised models are optimised for specific tasks such as code generation, image analysis or structured data processing. They offer an excellent performance/cost ratio for targeted use cases.
22 tokens/second
ministral-3:3b
Mistral AI's cutting-edge compact model, designed for efficiency in local and edge deployments.
Despite its small size, this model offers surprising performance for conversational tasks and simple reasoning. Ideal for mobile devices.
40 tokens/second
ministral-3:8b
Mid-sized model in the Ministral family, offering an optimal balance between performance and resources.
Version 8B is more robust, capable of handling longer contexts and more complex reasoning, while remaining very fast.
40 tokens/second
functiongemma:270m
Gemma micro-model specialising in function calling and detection of tool call intentions.
FunctionGemma 270M is an ultra-compact model optimised for identifying and formatting function calls. Ideal as a router or pre-filter in a multi-model agentic architecture.
49 tokens/second
granite3.2-vision:2b
IBM Granite compact multimodal model, specialising in the analysis of visual documents.
Granite 3.2 Vision 2B is a lightweight yet powerful model for OCR, data extraction from scanned documents and image analysis. Ideal for low-latency vision tasks.
qwen3-embedding:0.6b
Ultra-light Qwen3 embedding model, optimised for speed and efficiency on resource-limited infrastructures.
Offers an excellent compromise between semantic performance and speed of execution.
196.3 tokens/second
granite-embedding:278m
Ultra-compact IBM Granite embedding model, designed for maximum efficiency.
Ideal for semantic search tasks requiring minimal latency.
qwen3-embedding:4b
High-performance Qwen3-4B embedding model, offering deep semantic understanding and an extended context window.
Context of 40,000 tokens for processing large documents.
171 tokens/second
bge-m3:567m
State-of-the-art multilingual embedding model (BGE-M3), offering exceptional semantic search capabilities in over 100 languages.
Context of 8192 tokens. Supports dense, sparse and multi-vector search methods.
175 tokens/second
embeddinggemma:300m
Google's state-of-the-art embedding model, optimised for its size, ideal for search and semantic retrieval tasks.
Built on Gemma 3, this model produces vector representations of text for classification, clustering and similarity search. Trained on over 100 languages, its small size makes it perfect for resource-constrained environments.
57 tokens/second
gpt-oss:20b
OpenAI's open-weight language model, optimised for efficiency and deployment on consumer hardware.
A Mixture-of-Experts (MoE) model with 21 billion parameters and 3.6 billion active parameters. It offers configurable reasoning effort and agent capabilities.
55 tokens/second
qwen3-2507-think:4b
Qwen3-4B model optimised for reasoning, with improved performance on logic, maths, science and code tasks, and extended context to 250K tokens.
This 'Thinking' version has an increased thought length, making it ideal for highly complex reasoning tasks. It also offers general improvements in following instructions, using tools and generating text.
22 tokens/second
rnj-1:8b
Model 8B "Open Weight" specialising in coding, mathematics and science (STEM).
RNJ-1 is a dense model with 8.3B parameters trained on 8.4T tokens. It uses global attention and YaRN to provide a context of 32k tokens. It excels at code generation (83.5% HumanEval+) and mathematical reasoning, often outperforming much larger models.
64 tokens/second
qwen3-vl:2b
Ultra-compact multimodal Qwen3-VL model, bringing advanced vision capabilities to edge devices.
Despite its small size, this model incorporates Qwen3-VL technologies (MRoPE, DeepStack) to deliver impressive image and video analysis. Ideal for mobile or embedded applications requiring OCR, object detection or rapid visual understanding.
49 tokens/second
qwen3-vl:4b
Balanced Qwen3-VL multimodal model, offering robust vision performance with a small footprint.
Excellent compromise between performance and resources. Capable of analysing complex documents, graphics and videos with high accuracy. Supports structured extraction and visual reasoning.
16 tokens/second
qwen3.5:0.8b
Ultra-light Qwen3.5 model with 0.8 billion parameters, offering an exceptional native context of 250K tokens - a remarkable capacity for a model of this size.
Context configured to 250,000 tokens (native max context 262,144). Ideal for fast conversational tasks requiring a very long history or analysis of large documents with a small memory footprint.
37 tokens/second
qwen3.5:4b
Compact Qwen3.5 model with 4 billion parameters, offering a good compromise between performance and efficiency.
Context of 250k tokens. Good candidate for local assistants and light reasoning tasks.
32 tokens/second
qwen3.5:9b
Qwen3.5 model of intermediate size, offering solid reasoning capabilities with an extended context.
Context of 250k tokens. Offers a good balance between generation quality and inference speed.
46 tokens/second
qwen3:0.6b
Ultra-light Qwen3 model with 0.6 billion parameters, offering exceptional inference speed for fast, simple tasks.
Ideal for deployment on lightweight servers or as the first level of processing for complex workflows. Configured with a context of 40,000 tokens.
39 tokens/second
qwen3-vl:8b
Qwen3-VL multimodal model (8B), offering advanced vision performance with a reasonable footprint.
Version 8B of the Qwen3-VL model. Excellent compromise between performance and resources. Capable of analysing complex documents, graphics and video with high accuracy.
33 tokens/second
devstral-small-2:24b
Second iteration of Devstral (Small 2), a state-of-the-art agentic model for software engineering.
Optimised for codebase exploration, multi-file editing and tool use. Offers performance close to >100B models for code (SWE-bench Verified 68%). Native vision support. Context of 200k tokens.
84 tokens/second
deepseek-ocr
DeepSeek's specialist OCR model, designed for high-precision text extraction with formatting preservation.
Two-stage OCR system (visual encoder + MoE 3B decoder) optimised for converting documents into structured Markdown (tables, formulas). Requires specific pre-processing (Logits Processor) for optimum performance.
28 tokens/second
mistral-small3.2:24b
Minor update to Mistral Small 3.1, improving instruction tracking, function calling robustness and reducing repetition errors.
This version 3.2 retains the strengths of its predecessor while making targeted improvements. It is better able to follow precise instructions, produces fewer infinite generations or repetitive responses, and its function calling template is more robust.
27 tokens/second
translategemma:12b
State-of-the-art open translation model based on Gemma 3, covering 55 languages.
TranslateGemma 12B offers high-fidelity translation capabilities while respecting grammar and cultural nuances. Context of 128k tokens.
37 tokens/second
translategemma:4b
Compact version of the TranslateGemma translation model, optimised for speed.
TranslateGemma 4B offers fast and efficient translation capabilities for 55 languages. Context of 128k tokens.
16 tokens/second
translategemma:27b
High-performance translation model based on Gemma 3 27B.
TranslateGemma 27B offers superior translation quality for complex and technical content.
voxtral
Mistral AI's real-time ASR (Automatic Speech Recognition) model, capable of transcribing streaming audio via WebSocket.
Voxtral Mini 4B operates in Realtime mode via the /v1/realtime endpoint (WebSocket). It transcribes continuous audio with token extraction and ASR time tracking.
z-image:16b
Model for generating images from text prompts, compatible with the OpenAI /v1/images/generations API.
Z-Image Turbo is an image generation model compatible with the OpenAI Images API. It supports parameters for the size and number of images.
Model comparison
This comparison table will help you choose the model best suited to your needs, based on various criteria such as context size, performance and specific use cases.
| Model | Publisher | Parameters | Context (k tokens) | Vision | Agent | Reasoning | Security | Quick * | Energy efficiency * |
|---|---|---|---|---|---|---|---|---|---|
| Large models | |||||||||
| Specialised models | |||||||||
| ministral-3:3b | Mistral AI | 3B | 250000 | ||||||
| ministral-3:8b | Mistral AI | 8B | 250000 | ||||||
| functiongemma:270m | 270M | 32768 | |||||||
| granite3.2-vision:2b | IBM | 2B | 16384 | ||||||
| qwen3-embedding:0.6b | Qwen Team | 0.6B | 32768 | ||||||
| granite-embedding:278m | IBM | 278M | 512 | ||||||
| qwen3-embedding:4b | Qwen Team | 4B | 40000 | ||||||
| bge-m3:567m | BAAI | 567M | 8192 | ||||||
| embeddinggemma:300m | 300M | 2048 | |||||||
| gpt-oss:20b | OpenAI | 20B | 120000 | ||||||
| qwen3-2507-think:4b | Qwen Team | 4B | 250000 | ||||||
| rnj-1:8b | Essential AI | 8B | 32000 | ||||||
| qwen3-vl:2b | Qwen Team | 2B | 250000 | ||||||
| qwen3-vl:4b | Qwen Team | 4B | 250000 | ||||||
| qwen3.5:0.8b | Qwen Team | 0.8B | 250000 | ||||||
| qwen3.5:4b | Qwen Team | 4B | 250000 | ||||||
| qwen3.5:9b | Qwen Team | 9B | 250000 | ||||||
| qwen3:0.6b | Qwen Team | 0.6B | 40000 | ||||||
| qwen3-vl:8b | Qwen Team | 8B | 250000 | ||||||
| devstral-small-2:24b | Mistral AI & All Hands AI | 24B | 200000 | ||||||
| deepseek-ocr | DeepSeek AI | 3B | 8192 | ||||||
| mistral-small3.2:24b | Mistral AI | 24B | 128000 | ||||||
| translategemma:12b | 12B | 128000 | |||||||
| translategemma:4b | 4B | 128000 | |||||||
| translategemma:27b | 27B | 120000 | |||||||
| voxtral | Mistral AI | 4B | 32768 | N.C. | |||||
| z-image:16b | Community | 16B | N.C. | ||||||
Legend and explanation
Functionality or capacity supported by the model
Functionality or capability not supported by the model
* Energy efficiency
Indicates particularly low energy consumption (< 2.0 kWh/Mtoken)
* Quick
Model capable of generating more than 50 tokens per second
Note on performance measures
The speed values (tokens/s) represent performance targets in real-life conditions. Energy consumption (kWh/Mtoken) is calculated by dividing the estimated power of the inference server (in Watts) by the measured speed of the model (in tokens/second), then converted into kilowatt-hours per million tokens (division by 3.6). This method offers a practical comparison of the energy efficiency of different models, to be used as a relative indicator rather than an absolute measure of power consumption.
Recommended use cases
Here are some common use cases and the most suitable models for each. These recommendations are based on the specific performance and capabilities of each model.
Multilingual dialogue
Chatbots and assistants capable of communicating in several languages, with automatic detection, context maintenance throughout the conversation and understanding of linguistic specificities.
Recommended models
- nemotron-3-super:120b
- qwen3.5:27b
- nemotron3-nano:30b
- gpt-oss:120b
Analysis of long documents
Processing of large documents (>100 pages), maintaining context throughout the text, extracting key information, generating relevant summaries and answering specific content questions
Recommended models
- nemotron-3-super:120b
- qwen3.5:27b
- qwen3-2507:235b
Programming and development
Generating and optimising code in multiple languages, debugging, refactoring, developing complete functionalities, understanding complex algorithmic implementations and creating unit tests
Recommended models
- qwen3.5:27b
- qwen3-2507:235b
- qwen-coder-next:80b
- nemotron-3-super:120b
Visual analysis
Direct processing of images and visual documents without OCR pre-processing, interpretation of technical diagrams, graphs, tables, drawings and photos with generation of detailed textual explanations of the visual content
Recommended models
- qwen3.5:27b
- deepseek-ocr
- qwen3.5:35b
Safety and compliance
Applications requiring specific security capabilities; filtering of sensitive content, traceability of reasoning, RGPD/HDS verification, risk minimisation, vulnerability analysis and compliance with sectoral regulations
Recommended models
- granite3-guardian:8b
- qwen3.5:27b
- granite3-guardian:2b
Light and on-board deployments
Testing applications at Cloud Temple that require a minimal resource footprint, deployment on capacity-constrained devices, real-time inference on standard CPUs and integration into embedded or IoT systems.
Recommended models
- qwen3.5:0.8b
- qwen3-vl:2b
- ministral-3:3b
Contact us