Kagi Assistant's Android app fails to fully load AI answers when users leave the app while waiting, preventing follow-up questions without a full app restart, according to a user bug report. The user …
A user reported that searching "reasoning_effort" with the site:z.ai operator on Kagi returned 2 results from unrelated domains, github.com and aiproductivity.ai, when the Country/Region setting was s…
Kagi users on the company's KagiFeedback forum reported that the search engine's Quick Answer AI summaries returned brittle, slow, and sometimes false responses, including a hallucinated answer about …
Kagi stated that no user data goes to DeepSeek or Moonshot when its search assistant uses the Kimi and DeepSeek models, because Kagi runs those open-source models through independent third-party vendo…
Search engines Kagi and Google displayed OpenAI's March 17, 2026 GPT-5.4 announcement as "5 days ago," while Bing showed "17 hours ago" and Yandex showed "22 minutes ago," according to a report on the…
A Hacker News discussion flagged stack-junkie.com as AI slop appearing in Kagi's Small Web results, with user fxgn noting the site was added to the smallweb.txt list in a GitHub commit on August 6 and…
Mimo v2.6 Pro has taken the top spot on Artificial Analysis's price-to-performance Pareto frontier, according to a community post requesting that the model be added to a platform's standard plan. The …
Kagi's Quick Summary feature is intermittently failing to summarize Reddit results, according to a user report running Kagi on Comet on Windows 11. The bug affects some results on the same search resu…
A concept proposal argues that giving AI assistants access to the current time enables context-driven personalization, including time-windowed instruction profiles such as 06:00–09:00 for brief bullet…
A Kagi user reports an intermittent bug in which the search engine's summarization feature fails to feed webpage content to the underlying large language model, producing summaries without the page te…
A feature request on Kagi's feedback forum asks Kagi to add an "Ask" text-selection shortcut inside Kagi Assistant's own responses, letting users query or reply to a specific snippet without manually …
Kagi Search has claimed that AI models can perform up to 80% better when sourcing data using Kagi Search, but the company has not published an independent experiment to support the figure. The article…
Slash commands for an Assistant product remain undocumented and in an experimental state, according to a post that says the team has been on a detour fleshing out other parts of Assistant. The post st…
A feature request proposes displaying the cost of each query directly next to the finished answer in an AI Assistant, noting that the "AI usage" number is already updated immediately in the billing pa…
Users of GLM 5.3 Flash reported the model has become unusably slow, with response times of at least 3 minutes for simple questions, according to complaints posted on a discussion thread. Commenter shu…
Users of the Assistant product reported that reasoning models disappeared from the model picker, with one ultimate subscriber saying they could no longer select Sonnet with reasoning. A respondent sai…
Kagi Assistant fails to decode percent-encoded non-ASCII characters passed through the `q` URL parameter, so a query typed as `bügel` is submitted to the model as the literal string `b%C3%BCgel`. The …
A feature request proposes adding keyboard shortcuts to the Assistant for switching between AI models, such as Cmd/Ctrl + 1 for ChatGPT 4o and Cmd/Ctrl + 2 for Claude 3.5 Sonnet, matching the effect o…
Kagi users are requesting that the company's Assistant add an artifacts feature similar to Anthropic's Claude, which renders generated HTML, JavaScript, and SVG in a live side-by-side preview. Kagi te…
A Kagi user publicly asked the privacy-focused search company to clarify whether outputs from its AI features — Assistant, Universal Summarizer, FastGPT, and Translate — are watermarked or otherwise m…