An opiniontaed comparison, tips and Gotchas as of July 2026.
In July 2026, all LLM-based “AI” services are still prone to errors, hallucinations, low-quality code, and verbose botsplaining. Small energy-efficient models like Ecosia AI deliberately accept low quality output which might still be enough for many situations.
Anthropic’s Claude (Code) with its Sonnet, Haiku, Opus, and Fable models, is commonly seen as the best quality LLM AI especially for coding. However, Anthropic’s paid Claude Pro plan, currently for US$ 20,- per month, is often criticized as an upselling tool for the more expensive Claude (Pro) Max plan for 100,- dollars per month, which fails to offer a better value than the free tier.
All LLM AI providers make users pay for AI activity, measured in tokens, including erroneous output and wasted time. The free Claude web UI limits usage per day, while the paid Pro plan has a weekly limit. Many users report that the weekly limits is reached much too quickly, and they get a better value using one or multiple free Claude sessions instead. Changing web sessions means losing context, while on the CLI it might be possible to continue the same work with another free or paid plan after the first one has hit its limits.
Most other AI’s have similar usage limits and policies, including the French Le Chat by Mistral, Perplexity, and Copilot by Microsoft and GitHub.
Although mocked as the worst LLM AI by many users, Google Gemini and Google Search AI mode are also said to provide good enough quality answers for code analysis, vibe coding, bug fixing and chosing tools.
Open AI popularized AI with its free ChatGPT web app, also known as Chatty (or Shetty). The GPT model is still among the top performers for text operations in 2026, but the free version has limits now, and many users have refrained from using ChatGPT for ethical reasons, like OpenAI’s collaboration with the US military and secret service, resulting in an anti-OpenAI “quitGPT” movement comparable to the global “degoogle” initiative trying to ditch Google services for similar reasons.
Copilot lets its users choose LLM models, including GPT, Anthropic Claude, and Chinese DeepSeek AI models, but it defaults to GPT due to Microsoft’s investment in OpenAI. Despite good models and GitHub’s tooling, Copilot is often criticized for low quality, hallucinations, and too strict usage limits.
Le Chat is a French service using its own AI model, but the output sounds similar to GPT.
Thaura, Apertus, and Ecosia AI are other AI services apart from Chinese and US-American providers, focusing on the needs of European users concerned about privacy and ecological efficiency.