Some prompts stop when using stream execution. Let us know if you can reproduce the issue: prompt, session options, details about the device, etc. In the meantime, restart Chrome and try again. Last resort troubleshooting Uncaught NotSupportedError: … untested language … If you frequently encounter the following error, or if it appears with text that shouldn't be problematic, try disabling the Text Safety classifier in Chrome's flags settings: chrome://flags/#text-safety-classifier Uncaught NotSupportedError: The model attempted to output text in an untested language, and was prevented from doing so Alternative steps Some participants have reported that the following steps helped them get the component to show up: If you want to try the API on a device that doesn't have the performance or VRAM requirement (i.e., when capabilities() === 'no'), you can override it by setting the flag #optimization-guide-on-device-mode to "Enabled BypassPerfRequirement", then retry the setup steps. This comes with the caveat that, while the model may be able to run, it may also fail to execute with generic failure errors. Model download delay The browser may not start downloading the model right away. If your computer fulfills all the requirements and you don't see the model download start on chrome://components after calling LanguageModel.create(), and Optimization Guide On Device Model shows version 0.0.0.0 / New, leave the browser open for a few minutes to wait for the scheduler to start the download. Debug logs If everything fails: Open a new tab Go to chrome://gpu Download the gpu report Go to chrome://histograms/#OptimizationGuide.ModelExecution.OnDeviceModelInstallCriteria.AtRegistration.DiskSpace If you see records for 0, it means that your device doesn’t have enough storage space for the model. Ensure that you have at least 22 GB on the disk with your user profile, and retry the setup steps. If you are still stuck, continue with the other steps below.*On qualifying systems, the histogram should look similar to the following example:- Histogram: OptimizationGuide.ModelExecution.OnDeviceModelInstallCriteria.AtRegistration.DiskSpace recorded 1 samples, mean = 1.0 (flags = 0x41) [#] Go to chrome://histograms/#OptimizationGuide and download the histograms report Share both reports with the Early Preview Program coordinators. Crash logs If you encounter a crash error message such as “the model process crashed too many times for this version.”, then we’ll need a crash ID to investigate the root causes. Ensure that you have enabled crash reporting. Reproduce the issue. Go to chrome://crashes Find the most recent crash Hit Send now if necessary. Then wait and reload until the Status line changes from Not uploaded to Uploaded. Note that this could take a while. Copy the ID next to the Uploaded Crash Report ID line. Share the ID with the Early Preview Program coordinators. In the meantime, you’ll need to wait for a new Chrome version to get another chance of trying the Prompt API. Appendix Full API surface The full API surface is described below. See Web IDL for details on the language. The current implementation is a work in progress and does not support everything described below. [Exposed=(Window,Worker), SecureContext] interface LanguageModel : EventTarget { static Promise create(optional LanguageModelCreateOptions options = {}); static Promise availability(optional LanguageModelCreateCoreOptions options = {}); static Promise params(); // These will throw "NotSupportedError" DOMExceptions if role = "system" Promise prompt( LanguageModelPrompt input, optional LanguageModelPromptOptions options = {} ); ReadableStream promptStreaming( LanguageModelPrompt input, optional LanguageModelPromptOptions options = {} ); Promise append( LanguageModelPrompt input, optional LanguageModelAppendOptions options = {} ); Promise measureInputUsage( LanguageModelPrompt input, optional LanguageModelPromptOptions options = {} ); readonly attribute double inputUsage; readonly attribute unrestricted double inputQuota; attribute EventHandler onquotaoverflow; readonly attribute unsigned long topK; readonly attribute float temperature; Promise clone(optional LanguageModelCloneOptions options = {}); undefined destroy(); }; [Exposed=(Window,Worker), SecureContext] interface LanguageModelParams { readonly attribute unsigned long defaultTopK; readonly attribute unsigned long maxTopK; readonly attribute float defaultTemperature; readonly attribute float maxTemperature; }; dictionary LanguageModelCreateCoreOptions { // Note: these two have custom out-of-range handling behavior, not in the IDL layer. // They are unrestricted double so as to allow +Infinity without failing. unrestricted double topK; unrestricted double temperature; sequence expectedInputs; }; dictionary LanguageModelCreateOptions : LanguageModelCreateCoreOptions { AbortSignal signal; AICreateMonitorCallback monitor; LanguageModelInitialPrompts initialPrompts; }; dictionary LanguageModelPromptOptions { object responseConstraint; AbortSignal signal; }; dictionary LanguageModelAppendOptions { AbortSignal signal; }; dictionary LanguageModelCloneOptions { AbortSignal signal; }; dictionary LanguageModelExpectedInput { required LanguageModelMessageType type; sequence languages; }; // The argument to the prompt() method and others like it typedef ( // Canonical format sequence // Shorthand per the below comment or sequence // Shorthand for [{ role: "user", content: [{ type: "text", content: providedValue }] }] or DOMString ) LanguageModelPrompt; // The initialPrompts value omits the single string shorthand typedef ( // Canonical format sequence // Shorthand per the below comment or sequence ) LanguageModelInitialPrompts; dictionary LanguageModelMessage { required LanguageModelMessageRole role; required sequence content; }; // Shorthand for { role: providedValue.role, content: [{ type: "text", content: providedValue.content }] } dictionary LanguageModelMessageShorthand { required LanguageModelMessageRole role; required DOMString content; }; dictionary LanguageModelMessageContent { required LanguageModelMessageType type; required LanguageModelMessageContentValue content; }; dictionary LanguageModelPromptDict { LanguageModelMessageRole role = "user"; LanguageModelMessageType type = "text"; required LanguageModelMessageContent content; }; enum LanguageModelMessageRole { "system", "user", "assistant" }; enum LanguageModelMessageType { "text", "image", "audio" }; typedef ( ImageBitmapSource or AudioBuffer or BufferSource or DOMString ) LanguageModelMessageContentValue; 