The new descriptions of the different models in the Assistant have various mistakes. Below are the ones I could find, but I may have missed some. Additionally, some of these corrections may be debatable, and I may have made errors myself.
Claude 3.5 Haiku
"Anthropic's fast&cheap model. Multimodal. Currently superseded by alternatives like mistral small or Deepseek v3."
↓
"Anthropic's fast & cheap model. Multimodal. Currently superseded by alternatives like Mistral Small or DeepSeek V3."
All OpenAI models except o3 and o3 pro
The last sentence is missing a full stop.
All OpenAI models except o3, o3 pro, and o4 mini
"our users privacy" → "our users' privacy"
ChatGPT
"kagi" → "Kagi"
o3, o3 pro
"NOTE: this model can [...]" → "NOTE: This model can [...]"
Mistral Medium
"medium sized" → "medium-sized"
Mistral Large
">1year old" → "> 1 year old"
There are different opinions on spacing between ">" and "1". I've used a thin space in this case.
Gemini 2.5 Pro
"Reasoning+Multimodal (image&text)." → "Reasoning + multimodal (image & text)."
Llama 4 Maverick
"Medium sized" → "Medium-sized"
Qwen 3 235B
"Qwen-235B" → "Qwen3-235B" or "Qwen 3 235B"
DeepSeek R1 Distill Llama 70B
"Deepseek" → "DeepSeek"
Grok 4 Preview
"NOTE: The current rate limit is 60 requests per minute from xAI, expect frequent rate limiting."
↓
"NOTE: The current rate limit is 60 requests per minute from xAI; expect frequent rate limiting."
or
"NOTE: The current rate limit is 60 requests per minute from xAI. Expect frequent rate limiting."
Qwen 3 235B, Qwen 3 32B, DeepSeek R1, DeepSeek R1 Distill Llama 70B
Some models reference models by Alibaba, e.g. Qwen 3 32B's description includes "Qwen3-32B". While the names are consistent with the official models' names, they are different from the name in the dropdown menu, where dashes are omitted and spacing between "Qwen" and "3" is added.
