diff --git a/packages/kilo-vscode/tests/unit/search-match.test.ts b/packages/kilo-vscode/tests/unit/search-match.test.ts
new file mode 100644
index 00000000000..c12cf433abc
--- /dev/null
+++ b/packages/kilo-vscode/tests/unit/search-match.test.ts
@@ -0,0 +1,200 @@
+import { describe, it, expect } from "bun:test"
+import { acronymMatch, searchMatch } from "../../webview-ui/src/utils/search-match"
+
+// ---------------------------------------------------------------------------
+// acronymMatch — low-level word-boundary matching
+// ---------------------------------------------------------------------------
+
+describe("acronymMatch", () => {
+  describe("basic word boundary matching", () => {
+    it("matches at word start", () => {
+      expect(acronymMatch("fool org", "foo")).toBe(true)
+      expect(acronymMatch("the fool", "foo")).toBe(true)
+    })
+
+    it("does not match arbitrary substrings", () => {
+      expect(acronymMatch("faoboc", "foo")).toBe(false)
+      expect(acronymMatch("barfoo", "foo")).toBe(false)
+    })
+
+    it("matches prefix of a single word", () => {
+      expect(acronymMatch("foobar", "foob")).toBe(true)
+    })
+
+    it("matches exact word", () => {
+      expect(acronymMatch("test", "test")).toBe(true)
+      expect(acronymMatch("testing", "test")).toBe(true)
+      expect(acronymMatch("the test", "test")).toBe(true)
+    })
+  })
+
+  describe("word separators", () => {
+    it("recognizes space", () => {
+      expect(acronymMatch("hello world", "wor")).toBe(true)
+    })
+
+    it("recognizes hyphen", () => {
+      expect(acronymMatch("hello-world", "wor")).toBe(true)
+    })
+
+    it("recognizes underscore", () => {
+      expect(acronymMatch("hello_world", "wor")).toBe(true)
+    })
+
+    it("recognizes slash", () => {
+      expect(acronymMatch("hello/world", "wor")).toBe(true)
+    })
+
+    it("recognizes dot", () => {
+      expect(acronymMatch("hello.world", "wor")).toBe(true)
+    })
+
+    it("recognizes parentheses", () => {
+      expect(acronymMatch("Grok Code Fast 1 (free)", "free")).toBe(true)
+    })
+  })
+
+  describe("acronym matching", () => {
+    it("matches acronyms from word starts", () => {
+      expect(acronymMatch("Claude Sonnet", "clso")).toBe(true)
+    })
+
+    it("matches partial acronyms", () => {
+      expect(acronymMatch("Claude Sonnet 3.5", "cls")).toBe(true)
+    })
+
+    it("matches direct and acronym", () => {
+      expect(acronymMatch("clso tool", "clso")).toBe(true)
+      expect(acronymMatch("Claude Sonnet", "clso")).toBe(true)
+    })
+
+    it("does not match non-boundary acronym", () => {
+      expect(acronymMatch("aclbso", "clso")).toBe(false)
+    })
+  })
+
+  describe("camelCase and PascalCase", () => {
+    it("recognizes camelCase boundary", () => {
+      expect(acronymMatch("gitRebase", "gr")).toBe(true)
+    })
+
+    it("matches PascalCase acronyms", () => {
+      expect(acronymMatch("NewFileCreation", "nfc")).toBe(true)
+    })
+
+    it("splits camelCase at uppercase transitions", () => {
+      expect(acronymMatch("parseMarkdownContent", "pmc")).toBe(true)
+    })
+
+    it("handles mixed case scenarios", () => {
+      expect(acronymMatch("gitRebase", "gitr")).toBe(true)
+      expect(acronymMatch("GitRebase", "gitr")).toBe(true)
+    })
+  })
+
+  describe("backtracking", () => {
+    it("matches word that appears later in text", () => {
+      expect(acronymMatch("google gemini", "gemini")).toBe(true)
+      expect(acronymMatch("gemini pro", "gemini")).toBe(true)
+      expect(acronymMatch("google em emini", "gemini")).toBe(true)
+    })
+
+    it("matches partial word that appears later", () => {
+      expect(acronymMatch("Microsoft Copilot", "copilot")).toBe(true)
+      expect(acronymMatch("GitHub Copilot", "copilot")).toBe(true)
+    })
+
+    it("still respects word boundaries with backtracking", () => {
+      expect(acronymMatch("google gemini", "gemini")).toBe(true)
+      expect(acronymMatch("googlegemini", "gemini")).toBe(false)
+    })
+  })
+
+  describe("edge cases", () => {
+    it("handles empty text", () => {
+      expect(acronymMatch("", "foo")).toBe(false)
+    })
+
+    it("handles empty query", () => {
+      expect(acronymMatch("foo", "")).toBe(true)
+    })
+
+    it("handles special characters in text", () => {
+      expect(acronymMatch("foo-bar", "foob")).toBe(true)
+    })
+  })
+})
+
+// ---------------------------------------------------------------------------
+// searchMatch — high-level search with trimming and multi-word support
+// ---------------------------------------------------------------------------
+
+describe("searchMatch", () => {
+  describe("empty and whitespace queries", () => {
+    it("returns true for empty query", () => {
+      expect(searchMatch("", "anything")).toBe(true)
+    })
+
+    it("returns true for whitespace-only query", () => {
+      expect(searchMatch("   ", "anything")).toBe(true)
+    })
+  })
+
+  describe("case insensitivity", () => {
+    it("matches case-insensitively", () => {
+      expect(searchMatch("foo", "Foo Bar")).toBe(true)
+      expect(searchMatch("foo", "FOO BAZ")).toBe(true)
+      expect(searchMatch("FoO", "foo qux")).toBe(true)
+    })
+  })
+
+  describe("trimming", () => {
+    it("trims leading spaces", () => {
+      expect(searchMatch(" foo", "foo bar")).toBe(true)
+    })
+
+    it("trims trailing spaces", () => {
+      expect(searchMatch("foo ", "foo bar")).toBe(true)
+    })
+
+    it("trims spaces on both sides", () => {
+      expect(searchMatch("  foo  ", "foo bar")).toBe(true)
+    })
+  })
+
+  describe("multi-word queries", () => {
+    it("matches when all words present", () => {
+      expect(searchMatch("claude sonnet", "Claude Sonnet 3.5")).toBe(true)
+    })
+
+    it("does not match when any word missing", () => {
+      expect(searchMatch("claude sonnet", "Claude Opus")).toBe(false)
+      expect(searchMatch("claude sonnet", "GPT Sonnet")).toBe(false)
+    })
+  })
+
+  describe("real-world model search", () => {
+    it("finds model with hyphen in query", () => {
+      expect(searchMatch("gpt-5", "OpenAI: gpt-5 mini")).toBe(true)
+      expect(searchMatch("gpt-5", "OpenAI: gpt-4")).toBe(false)
+    })
+
+    it("finds model when hyphen omitted from query", () => {
+      expect(searchMatch("gpt5", "OpenAI: gpt-5 mini")).toBe(true)
+    })
+
+    it("finds all models with trailing hyphen", () => {
+      expect(searchMatch("gpt-", "OpenAI: gpt-5 mini")).toBe(true)
+      expect(searchMatch("gpt-", "OpenAI: gpt-4")).toBe(true)
+      expect(searchMatch("gpt-", "Anthropic: claude-3")).toBe(false)
+    })
+
+    it("matches file paths", () => {
+      expect(searchMatch("code", "src/services/code-index/manager.ts")).toBe(true)
+    })
+
+    it("matches mode selector options by label+value", () => {
+      expect(searchMatch("cod", "Code code Write code")).toBe(true)
+    })
+  })
+})
diff --git a/packages/kilo-vscode/webview-ui/src/components/shared/ModelSelector.tsx b/packages/kilo-vscode/webview-ui/src/components/shared/ModelSelector.tsx
index 1ae56f28f2b..dc7b8191361 100644
--- a/packages/kilo-vscode/webview-ui/src/components/shared/ModelSelector.tsx
+++ b/packages/kilo-vscode/webview-ui/src/components/shared/ModelSelector.tsx
@@ -29,6 +29,7 @@ import {
   sanitizeName,
 } from "./model-selector-utils"
 import { ModelPreview } from "./ModelPreview"
+import { searchMatch } from "../../utils/search-match"
 
 // ---------------------------------------------------------------------------
 // Row / group key helpers — single source of truth for key formatting
@@ -167,11 +168,13 @@ export const ModelSelectorBase: Component<ModelSelectorBaseProps> = (props) => {
 
   // Flat filtered list for keyboard navigation
   const filtered = createMemo(() => {
-    const q = debouncedSearch().toLowerCase()
+    const q = debouncedSearch().trim()
     if (!q) {
       return visibleModels()
     }
-    return visibleModels().filter((m) => m.name.toLowerCase().includes(q))
+    return visibleModels().filter(
+      (m) => searchMatch(q, m.name) || searchMatch(q, m.id) || searchMatch(q, m.providerName),
+    )
   })
 
   // Live set of favorited keys — drives star icon visual state (filled vs outline).
diff --git a/packages/kilo-vscode/webview-ui/src/utils/search-match.ts b/packages/kilo-vscode/webview-ui/src/utils/search-match.ts
new file mode 100644
index 00000000000..dfcc6ca0167
--- /dev/null
+++ b/packages/kilo-vscode/webview-ui/src/utils/search-match.ts
@@ -0,0 +1,59 @@
+/**
+ * Word-boundary matching for model/provider search.
+ *
+ * Splits text at camelCase transitions and common delimiters so that e.g.
+ * "clso" matches "Claude Sonnet" (Cl + So) and "gpt5" matches "gpt-5".
+ * Multi-word queries like "claude sonnet" require every fragment to match
+ * independently.
+ *
+ * Ported from legacy word-boundary-fzf.ts.
+ */
+
+// Split at positions before uppercase letters (camelCase/PascalCase)
+// and at common delimiters: hyphen, underscore, dot, colon, whitespace,
+// forward/back slash, brackets, parentheses.
+const WORD_BOUNDARY = /(?=[A-Z])|[[\]_.:\s/\\(){}-]+/
+
+/**
+ * Match a single query fragment against text using word-boundary acronym
+ * matching.  Each character in `query` must match the start of a word in
+ * `text`, consuming consecutive characters from the same word before moving
+ * to the next.
+ *
+ * Examples:
+ * - acronymMatch("Claude Sonnet", "clso") → true  (Cl + So)
+ * - acronymMatch("gitRebase", "gr") → true  (git + Rebase)
+ * - acronymMatch("faoboc", "foo") → false  (no word boundary)
+ */
+export function acronymMatch(text: string, query: string): boolean {
+  const words = text
+    .split(WORD_BOUNDARY)
+    .filter((w) => w.length > 0)
+    .map((w) => w.toLowerCase())
+
+  const attempt = (wi: number, qi: number): boolean => {
+    if (qi === query.length) return true
+    if (wi >= words.length) return false
+    const word = words[wi]!
+    let consumed = 0
+    while (qi + consumed < query.length && consumed < word.length && word[consumed] === query[qi + consumed]) consumed++
+    if (consumed > 0 && attempt(wi + 1, qi + consumed)) return true
+    return attempt(wi + 1, qi)
+  }
+
+  return attempt(0, 0)
+}
+
+/**
+ * High-level search: trims and lowercases the query, splits multi-word
+ * queries at word boundaries, and requires every fragment to match.
+ *
+ * Returns `true` when `query` is empty/whitespace-only.
+ */
+export function searchMatch(query: string, text: string): boolean {
+  const q = query.toLowerCase().trim()
+  if (!q) return true
+  const parts = q.split(WORD_BOUNDARY).filter((w) => w.length > 0)
+  if (parts.length === 0) return true
+  return parts.every((p) => acronymMatch(text, p))
+}