The government recently launched Compar: AI, a free platform to compare the answers of 24 chatbots. A good way to discover these famous artificial intelligence and develop your critical mind.
In a world where artificial intelligence is increasingly interfering in our daily lives, choosing the model most suited to your needs can be complex. Technology giants compete in ingenuity to offer ever more efficient AIs, making selection even more difficult for users. Chatgpt, Gemini, Meta Ai, Claude, Mistral … Which of them to choose? Which one is the most effective?
To help users make their choice and find the most useful AI, the government has developed Compar: ia. Created under the joint impetus of the Ministry of Culture and the Interministerial Directorate of Digital (Dinum), this entirely free platform aims to democratize access to the various conversational AIs by offering neophyte users the possibility of comparing them objectively and lit.
Compar: IA: a blind comparison platform
The platform gives access to the big names in artificial intelligence, some of which are usually paid. We find in particular GPT-4O of Openai, which constitutes one of the main engines of Chatgpt, but also Claude 3.5 Sonnet V2 of Anthropic, Gemini 2.0 Flash of Google, Deepseek V3, as well as Mistral Large, the big model of French language. Here is the full list of supported AI:
- COHERE/AYA EXPANSE 8B
- Anthropic/Claude 3.5 Sonnet V2
- COHERE/Command R
- Deepseek/Deepseek V3
- OPENAI/GPT-4O
- OPENAI/GPT-4O Mini
- Google/Gemini 1.5 Pro
- Google/Gemini 2.0 Flash
- Google/Gemma 2 9b
- We/Hermes 3 405B
- Liquid/LFM 40B
- META/LLAMA 3.1 405B
- Meta/Llama 3.1 70b
- META/LLAMA 3.1 8B
- NVIDIA/LLAMA 3.1 NEMOTRON 70B
- META/LLAMA 3.3 70B
- Mistral/Minister
- Mistral/Mistral Large
- Mistral/Mistral Nemo
- Mistral/Mistral Small 3
- Mistral/Mixtral 8x22b
- Microsoft/Phi-4
- Alibaba/Qwen2.5-Coder-32B
Compar: IA is accessible in beta version at This address. Its operation is simple: users are invited to interact simultaneously with two models of anonymized AI on subjects of their choice. At the end of each exchange, they express their preference for one of the responses, based on criteria such as relevance or utility. This participatory approach not only aims to guide users in their choice, but also to raise public awareness of the challenges of the generative AI and to encourage a critical mind in the face of the responses provided by these models – which can be completely false, especially during famous “hallucinations” – without being influenced by the names or the reputation of companies that have developed them.
Compar: IA: Develop the critical thinking of users
The data collected during these interactions is compiled in a database freely accessible to developers, researchers and model publishers. This information should help train new French -speaking AI models.
The platform also offers detailed exploration of the characteristics and conditions of use of the different models available. Note that it highlights the ecological dimension of technology, even if the measurement methodology remains in development. Thus, the electrical consumption of each model is estimated by taking into account several parameters, such as its size, the location of the servers, the carbon footprint linked to the manufacture of the equipment and the number of tokens generated during the responses. This data is then used to calculate its CO2 equivalent, so that users have the possibility of better understanding the impact of their research. An important issue that we tend to forget …