New version with significantly more hallucinations
The new version of Lumo can increasingly conduct better research. However, it is noticeable that language problems have persisted or even worsened. In German outputs, English is often unnecessarily mixed in; furthermore, spelling is frequently incorrect, resulting in misspelled words or distortions that alter their intended meaning. Occasionally, Chinese characters still appear in the output.
Additionally, it stands out that the model hallucinates significantly more than before and often uses web search only when prompted, even though it is now enabled by default. Positively, in comparison to other language models, there is no overly restrictive filter that triggers immediately with every statement. Topics that Gemini, for example, rejects solely due to certain keywords can be discussed here in the free version, though the security filters are presumably set too low.
Overall, the impression remains mixed: certain community requests have been addressed, but the overall quality has declined. Moreover, some content is possible that other LLMs deliberately exclude to prevent misuse. Ultimately, it must be stated that the temperature appears to be set too creatively, which weakens the model in some areas.