🌙

Kimi AI

Chatbots & Assistants
★★★★4.4/ 5 · note éditorialeFreemium
Publicité · 728×90

🗒️ Qu'est-ce que Kimi AI ?

L'assistant phare de Moonshot AI, propulsé par le modèle K2.5 à mille milliards de paramètres — codage agentique, recherche approfondie et chat quotidien pour une fraction du coût de l'API de GPT et Claude.

Kimi AI is built by Moonshot AI, a Beijing-based lab valued at roughly $18 billion, and has grown from a niche Chinese chatbot into a globally used assistant in under two years. Its K2.5 model uses a Mixture-of-Experts architecture with 1 trillion total parameters but activates only 32 billion per request, ranking competitively with GPT-5.4 and Claude Sonnet 4.6 on benchmarks while costing 4-17x less to run via the API. Kimi also offers a 300-subagent Agent Swarm mode and a desktop app that can execute tasks directly on your machine.

Kimi AI se distingue particulièrement par free tier includes unlimited basic chat, file uploads, and web search, k2.5 costs 4-17x less than GPT-5.4 and 5-6x less than Claude Sonnet 4.6 via the API for comparable benchmark performance et 300-subagent Agent Swarm mode for large parallel agentic tasks, ce qui en fait un choix populaire pour cost-sensitive teams running high-volume API workloads et agentic coding tasks via Kimi Code. À garder à l'esprit : built by a Chinese company, which may raise data-residency or compliance concerns for some organizations.

👍 Quels sont les avantages et inconvénients de Kimi AI ?

Avantages

  • Free tier includes unlimited basic chat, file uploads, and web search
  • K2.5 costs 4-17x less than GPT-5.4 and 5-6x less than Claude Sonnet 4.6 via the API for comparable benchmark performance
  • 300-subagent Agent Swarm mode for large parallel agentic tasks
  • Desktop app can execute tasks directly on your machine

Inconvénients

  • Built by a Chinese company, which may raise data-residency or compliance concerns for some organizations
  • Advanced agent features (Agent Swarm, Kimi Code, Kimi Claw) are locked behind paid tiers
  • Smaller ecosystem and track record than ChatGPT or Claude

🎯 À quoi sert Kimi AI ?

Cost-sensitive teams running high-volume API workloads
Agentic coding tasks via Kimi Code
Parallel research or automation tasks using Agent Swarm
General-purpose chat, file analysis, and web search

💰 Combien coûte Kimi AI ?

Kimi's chat app is free to use with unlimited basic chat, file uploads, and web search on its $0 Adagio plan; paid memberships range from $19 to $199 per month for heavier agentic and coding use. Via the API, Kimi K2.5 costs about $0.60 per 1M input tokens and $2.50-$3.00 per 1M output tokens -- 4 to 17x cheaper than GPT-5.4 and 5-6x cheaper than Claude Sonnet 4.6 for comparable benchmark performance.

Adagio (Free)

$0/mois
  • Unlimited basic chat, file uploads & web search
  • ~6 agent credits, 1 concurrent task
  • No Agent Swarm, Kimi Code, or Kimi Claw
Le plus populaire

Allegretto

$39/mois
  • 150 agent credits, 4 concurrent tasks
  • 50 Agent Swarm runs, Kimi Code at 5x speed
  • Kimi Claw browser control included

Vivace

$199/mois
  • 720 agent credits, 8 concurrent subtasks
  • 240 Agent Swarm runs
  • Kimi Code at 30x speed
Publicité · In-Content

Questions fréquentes

Kimi AI est-il gratuit ?

Kimi's chat app is free to use with unlimited basic chat, file uploads, and web search on its $0 Adagio plan; paid memberships range from $19 to $199 per month for heavier agentic and coding use. Via the API, Kimi K2.5 costs about $0.60 per 1M input tokens and $2.50-$3.00 per 1M output tokens -- 4 to 17x cheaper than GPT-5.4 and 5-6x cheaper than Claude Sonnet 4.6 for comparable benchmark performance.

À quoi sert Kimi AI ?

Kimi AI est couramment utilisé pour Cost-sensitive teams running high-volume API workloads, Agentic coding tasks via Kimi Code et Parallel research or automation tasks using Agent Swarm.

Kimi AI en vaut-il la peine en 2026 ?

Dans notre évaluation, Kimi AI obtient 4.4/5. Ses principaux atouts sont Free tier includes unlimited basic chat, file uploads, and web search et K2.5 costs 4-17x less than GPT-5.4 and 5-6x less than Claude Sonnet 4.6 via the API for comparable benchmark performance. En revanche, Built by a Chinese company, which may raise data-residency or compliance concerns for some organizations.

Avis des utilisateurs

Soyez le premier à laisser un avis !

Laisser un avis