diff --git a/docs/src/components/BuiltinTools.mdx b/docs/src/components/BuiltinTools.mdx
index 38052a00f4..e1eb2590c1 100644
--- a/docs/src/components/BuiltinTools.mdx
+++ b/docs/src/components/BuiltinTools.mdx
@@ -45,4 +45,5 @@ import { LinkCard } from '@astrojs/starlight/components';
 <LinkCard title="user_input_confirm" description="Ask the user to confirm a message." href="/genaiscript/reference/scripts/system#systemuser_input" />
 <LinkCard title="user_input_select" description="Ask the user to select an option." href="/genaiscript/reference/scripts/system#systemuser_input" />
 <LinkCard title="user_input_text" description="Ask the user to input text." href="/genaiscript/reference/scripts/system#systemuser_input" />
+<LinkCard title="vision_ask_image" description="Use vision model to run a query on an image" href="/genaiscript/reference/scripts/system#systemvision_ask_image" />
 
diff --git a/docs/src/content/docs/reference/cli/commands.md b/docs/src/content/docs/reference/cli/commands.md
index 5aa61c687e..500722c652 100644
--- a/docs/src/content/docs/reference/cli/commands.md
+++ b/docs/src/content/docs/reference/cli/commands.md
@@ -16,8 +16,9 @@ Usage: genaiscript run [options] <script> [files...]
 Runs a GenAIScript against files.
 
 Options:
-  -m, --model <string>                       model for the run
-  -sm, --small-model <string>                small model for the run
+  -m, --model <string>                       'large' model alias (default)
+  -sm, --small-model <string>                'small' alias model
+  -vm, --vision-model <string>               'vision' alias model
   -lp, --logprobs                            enable reporting token probabilities
   -tlp, --top-logprobs <number>              number of top logprobs (1 to 5)
   -ef, --excluded-files <string...>          excluded files
diff --git a/docs/src/content/docs/reference/scripts/system.mdx b/docs/src/content/docs/reference/scripts/system.mdx
index 6f5f060276..9c9df5448e 100644
--- a/docs/src/content/docs/reference/scripts/system.mdx
+++ b/docs/src/content/docs/reference/scripts/system.mdx
@@ -3350,6 +3350,72 @@ defTool(
 `````
 
 
+### `system.vision_ask_image`
+
+Vision Ask Image
+
+Register tool that uses vision model to run a query on an image
+
+-  tool `vision_ask_image`: Use vision model to run a query on an image
+
+`````js wrap title="system.vision_ask_image"
+system({
+    title: "Vision Ask Image",
+    description:
+        "Register tool that uses vision model to run a query on an image",
+})
+
+defTool(
+    "vision_ask_image",
+    "Use vision model to run a query on an image",
+    {
+        type: "object",
+        properties: {
+            image: {
+                type: "string",
+                description: "Image URL or workspace relative filepath",
+            },
+            query: {
+                type: "string",
+                description: "Query to run on the image",
+            },
+            hd: {
+                type: "boolean",
+                description: "Use high definition image",
+            },
+        },
+        required: ["image", "query"],
+    },
+    async (args) => {
+        const { image, query, hd } = args
+        const res = await runPrompt(
+            (_) => {
+                _.defImages(image, {
+                    autoCrop: true,
+                    detail: hd ? "high" : "low",
+                    maxWidth: hd ? 1024 : 512,
+                    maxHeight: hd ? 1024 : 512,
+                })
+                _.$`Answer this query about the images:`
+                _.def("QUERY", query)
+            },
+            {
+                model: "vision",
+                system: [
+                    "system",
+                    "system.assistant",
+                    "system.safety_jailbreak",
+                    "system.safety_harmful_content",
+                ],
+            }
+        )
+        return res
+    }
+)
+
+`````
+
+
 ### `system.zero_shot_cot`
 
 Zero-shot Chain Of Though
diff --git a/packages/cli/src/cli.ts b/packages/cli/src/cli.ts
index 5c083b4a8c..f5935648e1 100644
--- a/packages/cli/src/cli.ts
+++ b/packages/cli/src/cli.ts
@@ -96,8 +96,9 @@ export async function cli() {
         .command("run")
         .description("Runs a GenAIScript against files.")
         .arguments("<script> [files...]")
-        .option("-m, --model <string>", "model for the run")
-        .option("-sm, --small-model <string>", "small model for the run")
+        .option("-m, --model <string>", "'large' model alias (default)")
+        .option("-sm, --small-model <string>", "'small' alias model")
+        .option("-vm, --vision-model <string>", "'vision' alias model")
         .option("-lp, --logprobs", "enable reporting token probabilities")
         .option(
             "-tlp, --top-logprobs <number>",
diff --git a/packages/cli/src/nodehost.ts b/packages/cli/src/nodehost.ts
index d458f39280..3dd1bb428a 100644
--- a/packages/cli/src/nodehost.ts
+++ b/packages/cli/src/nodehost.ts
@@ -30,6 +30,7 @@ import {
     AZURE_AI_INFERENCE_TOKEN_SCOPES,
     MODEL_PROVIDER_AZURE_SERVERLESS_OPENAI,
     DOT_ENV_FILENAME,
+    DEFAULT_VISION_MODEL,
 } from "../../core/src/constants"
 import { tryReadText } from "../../core/src/fs"
 import {
@@ -141,6 +142,7 @@ export class NodeHost implements RuntimeHost {
     readonly defaultModelOptions = {
         model: DEFAULT_MODEL,
         smallModel: DEFAULT_SMALL_MODEL,
+        visionModel: DEFAULT_VISION_MODEL,
         temperature: DEFAULT_TEMPERATURE,
     }
     readonly defaultEmbeddingsModelOptions = {
diff --git a/packages/cli/src/run.ts b/packages/cli/src/run.ts
index 3d17e49754..7cf85c6db8 100644
--- a/packages/cli/src/run.ts
+++ b/packages/cli/src/run.ts
@@ -206,6 +206,8 @@ export async function runScript(
     if (options.model) host.defaultModelOptions.model = options.model
     if (options.smallModel)
         host.defaultModelOptions.smallModel = options.smallModel
+    if (options.visionModel)
+        host.defaultModelOptions.visionModel = options.visionModel
 
     const fail = (msg: string, exitCode: number, url?: string) => {
         logError(url ? `${msg} (see ${url})` : msg)
diff --git a/packages/cli/src/test.ts b/packages/cli/src/test.ts
index c6421e5b6a..852819c623 100644
--- a/packages/cli/src/test.ts
+++ b/packages/cli/src/test.ts
@@ -65,6 +65,7 @@ function parseModelSpec(m: string): ModelOptions {
         return {
             model: values["m"],
             smallModel: values["s"],
+            visionModel: values["v"],
             temperature: normalizeFloat(values["t"]),
             topP: normalizeFloat(values["p"]),
         }
@@ -120,11 +121,14 @@ export async function runPromptScriptTests(
         testDelay?: string
         model?: string
         smallModel?: string
+        visionModel?: string
     }
 ): Promise<PromptScriptTestRunResponse> {
     if (options.model) host.defaultModelOptions.model = options.model
     if (options.smallModel)
         host.defaultModelOptions.smallModel = options.smallModel
+    if (options.visionModel)
+        host.defaultModelOptions.visionModel = options.visionModel
 
     const scripts = await listTests({ ids, ...(options || {}) })
     if (!scripts.length)
@@ -163,6 +167,7 @@ export async function runPromptScriptTests(
             cli,
             model: info.model,
             smallModel: info.smallModel,
+            visionModel: info.visionModel,
             models: options.models?.map(parseModelSpec),
             provider: "provider.mjs",
             testProvider,
diff --git a/packages/core/src/chat.ts b/packages/core/src/chat.ts
index a196ef8188..016a03bd5f 100644
--- a/packages/core/src/chat.ts
+++ b/packages/core/src/chat.ts
@@ -345,11 +345,12 @@ ${fenceMD(content, " ")}
             logWarn(
                 `tool: ${tool.spec.name} response too long (${toolContentTokens} tokens), truncating ${maxToolContentTokens} tokens`
             )
-            toolContent = truncateTextToTokens(
-                toolContent,
-                maxToolContentTokens,
-                encoder
-            ) + "... (truncated)"
+            toolContent =
+                truncateTextToTokens(
+                    toolContent,
+                    maxToolContentTokens,
+                    encoder
+                ) + "... (truncated)"
         }
         trace.fence(toolContent, "markdown")
         toolResult.push(toolContent)
@@ -744,6 +745,10 @@ export function mergeGenerationOptions(
             runOptions?.smallModel ??
             options?.smallModel ??
             host.defaultModelOptions.smallModel,
+        visionModel:
+            runOptions?.visionModel ??
+            options?.visionModel ??
+            host.defaultModelOptions.visionModel,
         temperature:
             runOptions?.temperature ?? host.defaultModelOptions.temperature,
         embeddingsModel:
diff --git a/packages/core/src/connection.ts b/packages/core/src/connection.ts
index 4dc607f924..7b5e56c444 100644
--- a/packages/core/src/connection.ts
+++ b/packages/core/src/connection.ts
@@ -68,6 +68,9 @@ export async function parseDefaultsFromEnv(env: Record<string, string>) {
     if (env.GENAISCRIPT_DEFAULT_SMALL_MODEL)
         host.defaultModelOptions.smallModel =
             env.GENAISCRIPT_DEFAULT_SMALL_MODEL
+    if (env.GENAISCRIPT_DEFAULT_VISION_MODEL)
+        host.defaultModelOptions.visionModel =
+            env.GENAISCRIPT_DEFAULT_VISION_MODEL
     const t = normalizeFloat(env.GENAISCRIPT_DEFAULT_TEMPERATURE)
     if (!isNaN(t)) host.defaultModelOptions.temperature = t
     if (env.GENAISCRIPT_DEFAULT_EMBEDDINGS_MODEL)
diff --git a/packages/core/src/constants.ts b/packages/core/src/constants.ts
index a327c874ce..bd3ddf842c 100644
--- a/packages/core/src/constants.ts
+++ b/packages/core/src/constants.ts
@@ -53,6 +53,7 @@ export const RETRIEVAL_PERSIST_DIR = "retrieval"
 export const HIGHLIGHT_LENGTH = 4000
 export const SMALL_MODEL_ID = "small"
 export const LARGE_MODEL_ID = "large"
+export const VISION_MODEL_ID = "vision"
 export const DEFAULT_MODEL = "openai:gpt-4o"
 export const DEFAULT_MODEL_CANDIDATES = [
     "azure:gpt-4o",
@@ -62,6 +63,14 @@ export const DEFAULT_MODEL_CANDIDATES = [
     "github:gpt-4o",
     "client:gpt-4",
 ]
+export const DEFAULT_VISION_MODEL = "openai:gpt-4o"
+export const DEFAULT_VISION_MODEL_CANDIDATES = [
+    "azure:gpt-4o",
+    "azure_serverless:gpt-4o",
+    DEFAULT_MODEL,
+    "anthropic:claude-2",
+    "github:gpt-4o",
+]
 export const DEFAULT_SMALL_MODEL = "openai:gpt-4o-mini"
 export const DEFAULT_SMALL_MODEL_CANDIDATES = [
     "azure:gpt-4o-mini",
diff --git a/packages/core/src/genaiscript-api-provider.mjs b/packages/core/src/genaiscript-api-provider.mjs
index e951e21e07..1ffde3e5fa 100644
--- a/packages/core/src/genaiscript-api-provider.mjs
+++ b/packages/core/src/genaiscript-api-provider.mjs
@@ -22,8 +22,17 @@ class GenAIScriptApiProvider {
     }
 
     async callApi(prompt, context) {
-        const { model, smallModel, temperature, top_p, cache, version, cli, quiet } =
-            this.config
+        const {
+            model,
+            smallModel,
+            visionModel,
+            temperature,
+            top_p,
+            cache,
+            version,
+            cli,
+            quiet,
+        } = this.config
         const { vars, logger } = context
         try {
             let files = vars.files // string or string[]
@@ -52,6 +61,7 @@ class GenAIScriptApiProvider {
             if (quiet) args.push("--quiet")
             if (model) args.push("--model", model)
             if (smallModel) args.push("--small-model", smallModel)
+            if (visionModel) args.push("--vision-model", visionModel)
             if (temperature !== undefined)
                 args.push("--temperature", temperature)
             if (top_p !== undefined) args.push("--top_p", top_p)
diff --git a/packages/core/src/genaisrc/system.vision_ask_image.genai.mjs b/packages/core/src/genaisrc/system.vision_ask_image.genai.mjs
new file mode 100644
index 0000000000..cfff81b264
--- /dev/null
+++ b/packages/core/src/genaisrc/system.vision_ask_image.genai.mjs
@@ -0,0 +1,53 @@
+system({
+    title: "Vision Ask Image",
+    description:
+        "Register tool that uses vision model to run a query on an image",
+})
+
+defTool(
+    "vision_ask_image",
+    "Use vision model to run a query on an image",
+    {
+        type: "object",
+        properties: {
+            image: {
+                type: "string",
+                description: "Image URL or workspace relative filepath",
+            },
+            query: {
+                type: "string",
+                description: "Query to run on the image",
+            },
+            hd: {
+                type: "boolean",
+                description: "Use high definition image",
+            },
+        },
+        required: ["image", "query"],
+    },
+    async (args) => {
+        const { image, query, hd } = args
+        const res = await runPrompt(
+            (_) => {
+                _.defImages(image, {
+                    autoCrop: true,
+                    detail: hd ? "high" : "low",
+                    maxWidth: hd ? 1024 : 512,
+                    maxHeight: hd ? 1024 : 512,
+                })
+                _.$`Answer this query about the images:`
+                _.def("QUERY", query)
+            },
+            {
+                model: "vision",
+                system: [
+                    "system",
+                    "system.assistant",
+                    "system.safety_jailbreak",
+                    "system.safety_harmful_content",
+                ],
+            }
+        )
+        return res
+    }
+)
diff --git a/packages/core/src/host.ts b/packages/core/src/host.ts
index a1e5191cf6..5992ef7e16 100644
--- a/packages/core/src/host.ts
+++ b/packages/core/src/host.ts
@@ -125,7 +125,7 @@ export interface Host {
 
     // read a secret from the environment or a .env file
     defaultModelOptions: Required<
-        Pick<ModelOptions, "model" | "smallModel" | "temperature">
+        Pick<ModelOptions, "model" | "smallModel" | "visionModel" | "temperature">
     >
     defaultEmbeddingsModelOptions: Required<
         Pick<EmbeddingsModelOptions, "embeddingsModel">
diff --git a/packages/core/src/image.ts b/packages/core/src/image.ts
index fc347c2bac..4f1918acfb 100644
--- a/packages/core/src/image.ts
+++ b/packages/core/src/image.ts
@@ -15,11 +15,13 @@ export async function imageEncodeForLLM(
 ) {
     // Dynamically import the Jimp library and its alignment enums
     const { Jimp, HorizontalAlign, VerticalAlign } = await import("jimp")
-    const { autoCrop, maxHeight, maxWidth } = options
+    let { autoCrop, maxHeight, maxWidth } = options
 
     // If the URL is a string, resolve it to a data URI
     if (typeof url === "string") url = await resolveFileDataUri(url)
 
+    // https://platform.openai.com/docs/guides/vision/calculating-costs#managing-images
+
     // Return the URL if no image processing is required
     if (
         typeof url === "string" &&
diff --git a/packages/core/src/models.ts b/packages/core/src/models.ts
index 0fa1b925f2..e67c9cd57c 100644
--- a/packages/core/src/models.ts
+++ b/packages/core/src/models.ts
@@ -3,10 +3,12 @@ import {
     DEFAULT_EMBEDDINGS_MODEL_CANDIDATES,
     DEFAULT_MODEL_CANDIDATES,
     DEFAULT_SMALL_MODEL_CANDIDATES,
+    DEFAULT_VISION_MODEL_CANDIDATES,
     LARGE_MODEL_ID,
     MODEL_PROVIDER_LLAMAFILE,
     MODEL_PROVIDER_OPENAI,
     SMALL_MODEL_ID,
+    VISION_MODEL_ID,
 } from "./constants"
 import { errorMessage } from "./error"
 import { LanguageModelConfiguration, host } from "./host"
@@ -107,6 +109,12 @@ export async function resolveModelConnectionInfo(
             host.defaultModelOptions.smallModel,
             ...DEFAULT_SMALL_MODEL_CANDIDATES,
         ]
+    } else if (m === VISION_MODEL_ID) {
+        m = undefined
+        candidates ??= [
+            host.defaultModelOptions.visionModel,
+            ...DEFAULT_VISION_MODEL_CANDIDATES,
+        ]
     } else if (m === LARGE_MODEL_ID) {
         m = undefined
         candidates ??= [
diff --git a/packages/core/src/runpromptcontext.ts b/packages/core/src/runpromptcontext.ts
index 8356d1e43f..66f76695d2 100644
--- a/packages/core/src/runpromptcontext.ts
+++ b/packages/core/src/runpromptcontext.ts
@@ -19,6 +19,7 @@ import {
     createFileMerge,
     createSystemNode,
     finalizeMessages,
+    PromptImage,
 } from "./promptdom"
 import { MarkdownTrace } from "./trace"
 import { GenerationOptions } from "./generation"
@@ -39,6 +40,7 @@ import {
     appendSystemMessage,
     executeChatSession,
     mergeGenerationOptions,
+    toChatCompletionUserMessage,
     tracePromptResult,
 } from "./chat"
 import { checkCancelled } from "./cancellation"
@@ -621,6 +623,7 @@ export function createChatGenerationContext(
             let tools: ToolCallback[] = undefined
             let schemas: Record<string, JSONSchema> = undefined
             let chatParticipants: ChatParticipant[] = undefined
+            const images: PromptImage[] = []
             const fileMerges: FileMergeHandler[] = []
             const outputProcessors: PromptOutputProcessorHandler[] = []
             const fileOutputs: FileOutput[] = []
@@ -641,6 +644,7 @@ export function createChatGenerationContext(
                     fileMerges: fms,
                     outputProcessors: ops,
                     fileOutputs: fos,
+                    images: imgs,
                 } = await renderPromptNode(genOptions.model, node, {
                     flexTokens: genOptions.flexTokens,
                     trace: runTrace,
@@ -653,6 +657,7 @@ export function createChatGenerationContext(
                 fileMerges.push(...fms)
                 outputProcessors.push(...ops)
                 fileOutputs.push(...fos)
+                images.push(...imgs)
 
                 if (errors?.length) {
                     logError(errors.map((err) => errorMessage(err)).join("\n"))
@@ -731,6 +736,9 @@ export function createChatGenerationContext(
                 addToolDefinitionsMessage(messages, tools)
                 genOptions.fallbackTools = true
             }
+            if (images?.length)
+                messages.push(toChatCompletionUserMessage("", images))
+
             finalizeMessages(messages, { fileOutputs })
             const connection = await resolveModelConnectionInfo(genOptions, {
                 trace: runTrace,
diff --git a/packages/core/src/server/messages.ts b/packages/core/src/server/messages.ts
index b39af4cfdf..79b24d2f25 100644
--- a/packages/core/src/server/messages.ts
+++ b/packages/core/src/server/messages.ts
@@ -70,6 +70,7 @@ export interface PromptScriptRunOptions {
     maxDataRepairs: string
     model: string
     smallModel: string
+    visionModel: string
     embeddingsModel: string
     csvSeparator: string
     cache: boolean | string
diff --git a/packages/core/src/test.ts b/packages/core/src/test.ts
index b127d01741..a15ae4bb72 100644
--- a/packages/core/src/test.ts
+++ b/packages/core/src/test.ts
@@ -1,5 +1,10 @@
 // Import necessary utilities and constants
-import { HTTPS_REGEX, LARGE_MODEL_ID, SMALL_MODEL_ID } from "./constants"
+import {
+    HTTPS_REGEX,
+    LARGE_MODEL_ID,
+    SMALL_MODEL_ID,
+    VISION_MODEL_ID,
+} from "./constants"
 import { arrayify } from "./util"
 import { host } from "./host"
 
@@ -40,6 +45,7 @@ export function generatePromptFooConfiguration(
         cli?: string
         model?: string
         smallModel?: string
+        visionModel?: string
         models?: ModelOptions[]
     }
 ) {
@@ -54,6 +60,7 @@ export function generatePromptFooConfiguration(
             ...script,
             model: options?.model || script.model,
             smallModel: options?.smallModel || script.smallModel,
+            visionModel: options?.visionModel || script.visionModel,
         })
     }
 
@@ -63,9 +70,11 @@ export function generatePromptFooConfiguration(
     const resolveModel = (m: string) =>
         m === SMALL_MODEL_ID
             ? host.defaultModelOptions.smallModel
-            : m === LARGE_MODEL_ID
-              ? host.defaultModelOptions.model
-              : m
+            : m === VISION_MODEL_ID
+              ? host.defaultModelOptions.visionModel
+              : m === LARGE_MODEL_ID
+                ? host.defaultModelOptions.model
+                : m
 
     // Create configuration object
     const res = {
@@ -74,21 +83,25 @@ export function generatePromptFooConfiguration(
         prompts: [id],
         // Map model options to providers
         providers: models
-            .map(({ model, smallModel, temperature, topP }) => ({
+            .map(({ model, smallModel, visionModel, temperature, topP }) => ({
                 model: resolveModel(model) ?? host.defaultModelOptions.model,
                 smallModel:
                     resolveModel(smallModel) ??
                     host.defaultModelOptions.smallModel,
+                visionModel:
+                    resolveModel(visionModel) ??
+                    host.defaultModelOptions.visionModel,
                 temperature: !isNaN(temperature)
                     ? temperature
                     : host.defaultModelOptions.temperature,
                 top_p: topP,
             }))
-            .map(({ model, smallModel, temperature, top_p }) => ({
+            .map(({ model, smallModel, visionModel, temperature, top_p }) => ({
                 id: provider,
                 label: [
                     model,
                     smallModel,
+                    visionModel,
                     `t=${temperature}`,
                     top_p !== undefined ? `p=${top_p}` : undefined,
                 ]
@@ -97,6 +110,7 @@ export function generatePromptFooConfiguration(
                 config: {
                     model,
                     smallModel,
+                    visionModel,
                     temperature,
                     top_p,
                     cli,
diff --git a/packages/core/src/testhost.ts b/packages/core/src/testhost.ts
index 38e284c9fe..ecae750b57 100644
--- a/packages/core/src/testhost.ts
+++ b/packages/core/src/testhost.ts
@@ -22,6 +22,7 @@ import {
     DEFAULT_MODEL,
     DEFAULT_SMALL_MODEL,
     DEFAULT_TEMPERATURE,
+    DEFAULT_VISION_MODEL,
 } from "./constants"
 import {
     dirname,
@@ -73,6 +74,7 @@ export class TestHost implements RuntimeHost {
     readonly defaultModelOptions = {
         model: DEFAULT_MODEL,
         smallModel: DEFAULT_SMALL_MODEL,
+        visionModel: DEFAULT_VISION_MODEL,
         temperature: DEFAULT_TEMPERATURE,
     }
     // Default options for embeddings models
diff --git a/packages/core/src/types/prompt_template.d.ts b/packages/core/src/types/prompt_template.d.ts
index 04e315f981..da3e818363 100644
--- a/packages/core/src/types/prompt_template.d.ts
+++ b/packages/core/src/types/prompt_template.d.ts
@@ -121,6 +121,7 @@ type PromptTemplateResponseType = "json_object" | "json_schema" | undefined
 type ModelType = OptionsOrString<
     | "large"
     | "small"
+    | "vision"
     | "openai:gpt-4o"
     | "openai:gpt-4o-mini"
     | "openai:gpt-3.5-turbo"
@@ -156,6 +157,10 @@ type ModelSmallType = OptionsOrString<
     | "github:AI21-Jamba-1-5-Mini"
 >
 
+type ModelVisionType = OptionsOrString<
+    "openai:gpt-4o" | "github:gpt-4o" | "azure:gpt-4o" | "azure:gpt-4o-mini"
+>
+
 interface ModelConnectionOptions {
     /**
      * Which LLM model to use. Use `large` for the default set of model candidates, `small` for the set of small models like gpt-4o-mini.
@@ -169,6 +174,11 @@ interface ModelConnectionOptions {
      * @example gpt-4
      */
     smallModel?: ModelSmallType
+
+    /**
+     * Which LLM to use for the "vision" model.
+     */
+    visionModel?: ModelVisionType
 }
 
 interface ModelOptions extends ModelConnectionOptions {
@@ -782,7 +792,7 @@ interface ExpansionVariables {
         /**
          * When running in GitHub Copilot Chat, the current terminal content
          */
-        "copilot.terminalSelection?": string
+        "copilot.terminalSelection"?: string
     }
 
     /**
diff --git a/packages/sample/genaisrc/ask-image-tool.genai.mjs b/packages/sample/genaisrc/ask-image-tool.genai.mjs
new file mode 100644
index 0000000000..e672ff18bb
--- /dev/null
+++ b/packages/sample/genaisrc/ask-image-tool.genai.mjs
@@ -0,0 +1,5 @@
+script({
+    tools: "vision_ask_image"
+})
+
+$`What kind of objects are on src/robots.jpg?`
\ No newline at end of file
diff --git a/packages/sample/genaisrc/defimages-inline.genai.mjs b/packages/sample/genaisrc/defimages-inline.genai.mjs
new file mode 100644
index 0000000000..af43c5b1dc
--- /dev/null
+++ b/packages/sample/genaisrc/defimages-inline.genai.mjs
@@ -0,0 +1,8 @@
+script({
+    files: "src/robots.jpg",
+    model: "github:gpt-4o",
+})
+const res = await runPrompt((_) => {
+    _.defImages(env.files)
+    _.$`give keywords describing for each image`
+})
diff --git a/packages/vscode/src/vshost.ts b/packages/vscode/src/vshost.ts
index 44db4ee725..91d1bbab3e 100644
--- a/packages/vscode/src/vshost.ts
+++ b/packages/vscode/src/vshost.ts
@@ -12,6 +12,7 @@ import {
     DEFAULT_MODEL,
     DEFAULT_SMALL_MODEL,
     DEFAULT_TEMPERATURE,
+    DEFAULT_VISION_MODEL,
 } from "../../core/src/constants"
 import { dotEnvTryParse } from "../../core/src/dotenv"
 import {
@@ -32,6 +33,7 @@ export class VSCodeHost extends EventTarget implements Host {
     readonly defaultModelOptions = {
         model: DEFAULT_MODEL,
         smallModel: DEFAULT_SMALL_MODEL,
+        visionModel: DEFAULT_VISION_MODEL,
         temperature: DEFAULT_TEMPERATURE,
     }
     readonly defaultEmbeddingsModelOptions = {