diff --git a/next.config.mjs b/next.config.mjs
index 3853b4b0..e71a530a 100644
--- a/next.config.mjs
+++ b/next.config.mjs
@@ -28,6 +28,8 @@ const nextConfig = {
   experimental: {
     // #1529/#1572: LLM clients can send long context or base64 image payloads through /v1 rewrites.
     proxyClientMaxBodySize,
+    // Cache fetch responses across HMR refreshes for faster dev reloads.
+    serverComponentsHmrCache: true,
   },
   webpack: (config, { isServer }) => {
     // Ignore fs/path modules in browser bundle
diff --git a/open-sse/providers/capabilities.js b/open-sse/providers/capabilities.js
index fe9fcbb5..5f04ec23 100644
--- a/open-sse/providers/capabilities.js
+++ b/open-sse/providers/capabilities.js
@@ -158,7 +158,8 @@ export const PATTERN_CAPABILITIES = [
   { pattern: "*deepseek-v4*",   caps: { reasoning: true, thinkingFormat: "deepseek", contextWindow: 1000000, maxOutput: 384000 } },
   { pattern: "*reasoner*",      caps: { reasoning: true, thinkingFormat: "deepseek", thinkingCanDisable: false, contextWindow: 128000 } },
   { pattern: "*deepseek-r*",    caps: { reasoning: true, thinkingFormat: "deepseek", thinkingCanDisable: false, contextWindow: 128000 } },
-  { pattern: "*deepseek*",      caps: { contextWindow: 128000 } },
+  { pattern: "*deepseek-chat*", caps: { contextWindow: 128000 } },
+  { pattern: "*deepseek*",      caps: { reasoning: true, thinkingFormat: "deepseek", contextWindow: 128000 } },
 
   // ── MiniMax (M3 = adaptive; M2.x cannot disable) ─────────────────
   { pattern: "*minimax*image*", caps: { imageOutput: true } },
diff --git a/open-sse/providers/registry/gemini-cli.js b/open-sse/providers/registry/gemini-cli.js
index 8af51690..3e4a94c2 100644
--- a/open-sse/providers/registry/gemini-cli.js
+++ b/open-sse/providers/registry/gemini-cli.js
@@ -31,8 +31,13 @@ export default {
     clientSecret: "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",
   },
   models: [
-    { id: "gemini-3-flash-preview", name: "Gemini 3 Flash Preview" },
+    { id: "gemini-3.1-pro-preview", name: "Gemini 3.1 Pro Preview" },
     { id: "gemini-3-pro-preview", name: "Gemini 3 Pro Preview" },
+    { id: "gemini-3-flash-preview", name: "Gemini 3 Flash Preview" },
+    { id: "gemini-3.1-flash-lite-preview", name: "Gemini 3.1 Flash Lite Preview" },
+    { id: "gemini-2.5-pro", name: "Gemini 2.5 Pro" },
+    { id: "gemini-2.5-flash", name: "Gemini 2.5 Flash" },
+    { id: "gemini-2.5-flash-lite", name: "Gemini 2.5 Flash Lite" },
   ],
   oauth: {
     authorizeUrl: "https://accounts.google.com/o/oauth2/v2/auth",
diff --git a/open-sse/providers/registry/github.js b/open-sse/providers/registry/github.js
index 9251c5c2..95169eb3 100644
--- a/open-sse/providers/registry/github.js
+++ b/open-sse/providers/registry/github.js
@@ -41,12 +41,6 @@ export default {
     },
   },
   models: [
-    { id: "gpt-3.5-turbo", name: "GPT-3.5 Turbo" },
-    { id: "gpt-4", name: "GPT-4" },
-    { id: "gpt-4o", name: "GPT-4o" },
-    { id: "gpt-4o-mini", name: "GPT-4o mini" },
-    { id: "gpt-4.1", name: "GPT-4.1" },
-    { id: "gpt-5-mini", name: "GPT-5 Mini" },
     { id: "gpt-5.2", name: "GPT-5.2" },
     { id: "gpt-5.2-codex", name: "GPT-5.2 Codex" },
     { id: "gpt-5.3-codex", name: "GPT-5.3 Codex" },
@@ -54,7 +48,6 @@ export default {
     { id: "gpt-5.4-mini", name: "GPT-5.4 Mini" },
     { id: "claude-haiku-4.5", name: "Claude Haiku 4.5" },
     { id: "claude-opus-4.5", name: "Claude Opus 4.5" },
-    { id: "claude-sonnet-4", name: "Claude Sonnet 4" },
     { id: "claude-sonnet-4.5", name: "Claude Sonnet 4.5" },
     { id: "claude-sonnet-4.6", name: "Claude Sonnet 4.6" },
     { id: "claude-opus-4.6", name: "Claude Opus 4.6" },
diff --git a/open-sse/providers/registry/glm-cn.js b/open-sse/providers/registry/glm-cn.js
index c19b4fed..90a71b4c 100644
--- a/open-sse/providers/registry/glm-cn.js
+++ b/open-sse/providers/registry/glm-cn.js
@@ -21,6 +21,7 @@ export default {
     },
   },
   models: [
+    { id: "glm-5.2", name: "GLM 5.2" },
     { id: "glm-5.1", name: "GLM 5.1" },
     { id: "glm-5", name: "GLM 5" },
     { id: "glm-4.7", name: "GLM-4.7" },
diff --git a/open-sse/providers/registry/glm.js b/open-sse/providers/registry/glm.js
index fa003ccf..ceeb8796 100644
--- a/open-sse/providers/registry/glm.js
+++ b/open-sse/providers/registry/glm.js
@@ -33,6 +33,7 @@ export default {
     },
   },
   models: [
+    { id: "glm-5.2", name: "GLM 5.2" },
     { id: "glm-5.1", name: "GLM 5.1" },
     { id: "glm-5", name: "GLM 5" },
     { id: "glm-4.7", name: "GLM 4.7" },
diff --git a/open-sse/services/combo.js b/open-sse/services/combo.js
index deaa0dc1..2719d7c2 100644
--- a/open-sse/services/combo.js
+++ b/open-sse/services/combo.js
@@ -39,15 +39,16 @@ export function reorderByCapabilities(models, required) {
  */
 const comboRotationState = new Map();
 
-// Last array item whose role is "user" (current turn), or the last item when no
-// role is present. History media (older turns) must not pin the combo to a vision
-// model — those get stripped + placeholdered downstream instead.
-function lastUserItem(arr) {
-  if (!Array.isArray(arr) || arr.length === 0) return null;
-  for (let i = arr.length - 1; i >= 0; i--) {
-    if (!arr[i]?.role || arr[i].role === "user") return arr[i];
-  }
-  return arr[arr.length - 1];
+// Trailing run of items after the last assistant/model turn = the current user
+// turn. It may span several messages (e.g. text + image split across blocks),
+// so we return all of them. History media (older turns) must not pin the combo
+// to a vision model — those get stripped + placeholdered downstream instead.
+function trailingUserItems(arr) {
+  if (!Array.isArray(arr) || arr.length === 0) return [];
+  const isAssistant = (r) => r === "assistant" || r === "model";
+  let i = arr.length - 1;
+  while (i >= 0 && !isAssistant(arr[i]?.role)) i--;
+  return arr.slice(i + 1);
 }
 
 // Detect which capabilities a request needs. Modalities (vision/pdf) are scanned
@@ -72,14 +73,11 @@ export function detectRequiredCapabilities(body) {
     if (Array.isArray(content)) for (const b of content) scanBlock(b);
   };
 
-  // Modalities: current user turn only (last item across each known shape).
-  const lastMsg = lastUserItem(body.messages);     // openai / claude
-  if (lastMsg) scanContent(lastMsg.content);
-  const lastInput = lastUserItem(body.input);      // responses
-  if (lastInput) scanContent(lastInput.content);
-  const contents = body.contents || body.request?.contents; // gemini / antigravity
-  const lastContent = lastUserItem(contents);
-  if (lastContent) scanContent(lastContent.parts);
+  // Modalities: current user turn only (trailing user run across each known shape).
+  for (const m of trailingUserItems(body.messages)) scanContent(m.content);      // openai / claude
+  for (const it of trailingUserItems(body.input)) scanContent(it.content);       // responses
+  const contents = body.contents || body.request?.contents;                      // gemini / antigravity
+  for (const c of trailingUserItems(contents)) scanContent(c.parts);
 
   // search: temporarily disabled in auto-switch (feature not wired yet).
 
diff --git a/open-sse/translator/concerns/thinkingUnified.js b/open-sse/translator/concerns/thinkingUnified.js
index 1f110a2f..0ed27608 100644
--- a/open-sse/translator/concerns/thinkingUnified.js
+++ b/open-sse/translator/concerns/thinkingUnified.js
@@ -40,6 +40,15 @@ export function parseSuffix(model) {
 export function extractThinking(body) {
   if (!body || typeof body !== "object") return null;
 
+  // Claude output_config.effort (explicit) — priority over adaptive thinking
+  const oc = body.output_config?.effort;
+  if (typeof oc === "string" && oc) {
+    const e = oc.toLowerCase();
+    if (e === "none" || e === "off") return { mode: "none" };
+    if (e === "auto") return { mode: "auto" };
+    return { mode: "level", level: e };
+  }
+
   // Claude shape
   const t = body.thinking;
   if (t && typeof t === "object") {
@@ -154,7 +163,7 @@ function applyFormat(fmt, body, cfg, caps) {
     case "openai": {
       if (none && canDisable) { body.reasoning_effort = "none"; break; }
       const level = toLevel(eff);
-      if (level) body.reasoning_effort = level;
+      if (level) body.reasoning_effort = level === "xhigh" || level === "max" ? "high" : level;
       break;
     }
     case "claude-adaptive": {
diff --git a/open-sse/translator/formats/claude.js b/open-sse/translator/formats/claude.js
index 24c6a5fe..1dac5c33 100644
--- a/open-sse/translator/formats/claude.js
+++ b/open-sse/translator/formats/claude.js
@@ -5,6 +5,8 @@ import { adjustMaxTokens } from "./maxTokens.js";
 import { applyCloaking } from "../../utils/claudeCloaking.js";
 import { resolveSessionId } from "../../utils/sessionManager.js";
 import { PROVIDERS } from "../../providers/index.js";
+import { getCapabilitiesForModel } from "../../providers/capabilities.js";
+import { DEFAULT_MAX_TOKENS } from "../../config/runtimeConfig.js";
 
 // Check if message has valid non-empty content
 export function hasValidContent(msg) {
@@ -130,12 +132,18 @@ export function normalizeClaudePassthrough(body, model = "") {
 // - Add thinking block for Anthropic endpoint (provider === "claude")
 // - Fix tool_use/tool_result ordering
 // - Apply cloaking (billing header + fake user ID) for OAuth tokens
-export function prepareClaudeRequest(body, provider = null, apiKey = null, connectionId = null) {
+export function prepareClaudeRequest(body, provider = null, apiKey = null, connectionId = null, rawHeaders = null, sessionId = null) {
   // quirk: MiniMax's Claude-compatible endpoint rejects Anthropic's output_config (400 invalid params)
   if (PROVIDERS[provider]?.quirks?.dropOutputConfig) {
     delete body.output_config;
   }
 
+  // Clamp max_tokens to the model output ceiling (never above DEFAULT_MAX_TOKENS)
+  if (body.max_tokens) {
+    const ceiling = Math.min(getCapabilitiesForModel(provider, body.model).maxOutput, DEFAULT_MAX_TOKENS);
+    if (body.max_tokens > ceiling) body.max_tokens = ceiling;
+  }
+
   // 1. System: remove all cache_control, add only to last block with ttl 1h
   if (body.system && Array.isArray(body.system)) {
     body.system = body.system.map((block, i) => {
@@ -252,8 +260,8 @@ export function prepareClaudeRequest(body, provider = null, apiKey = null, conne
   // Apply cloaking for OAuth tokens (billing header + fake user ID)
   // session_id in user_id must match X-Claude-Code-Session-Id for fingerprint consistency
   if ((provider === "claude" || provider?.startsWith("anthropic-compatible")) && apiKey) {
-    const sessionId = resolveSessionId({ body, connectionId, scope: "claude" });
-    body = applyCloaking(body, apiKey, sessionId);
+    const sid = sessionId || resolveSessionId({ headers: rawHeaders, body, connectionId, scope: "claude" });
+    body = applyCloaking(body, apiKey, sid);
   }
 
   return body;
diff --git a/open-sse/translator/formats/maxTokens.js b/open-sse/translator/formats/maxTokens.js
index 294d9c55..0e5b36f2 100644
--- a/open-sse/translator/formats/maxTokens.js
+++ b/open-sse/translator/formats/maxTokens.js
@@ -8,7 +8,7 @@ import { DEFAULT_MAX_TOKENS, DEFAULT_MIN_TOKENS } from "../../config/runtimeConf
 export function adjustMaxTokens(body) {
   let maxTokens = body.max_tokens || DEFAULT_MAX_TOKENS;
 
-  // Auto-increase for tool calling to prevent truncated arguments
+  // Auto-increase for tool calling to prevent truncated arguments (min never above max)
   if (body.tools && Array.isArray(body.tools) && body.tools.length > 0) {
     if (maxTokens < DEFAULT_MIN_TOKENS) {
       maxTokens = DEFAULT_MIN_TOKENS;
@@ -22,6 +22,9 @@ export function adjustMaxTokens(body) {
     maxTokens = body.thinking.budget_tokens + 1024;
   }
 
+  // Never exceed the global ceiling
+  if (maxTokens > DEFAULT_MAX_TOKENS) maxTokens = DEFAULT_MAX_TOKENS;
+
   return maxTokens;
 }
 
diff --git a/open-sse/translator/index.js b/open-sse/translator/index.js
index fa42b139..5c9065d2 100644
--- a/open-sse/translator/index.js
+++ b/open-sse/translator/index.js
@@ -5,6 +5,7 @@ import { cloakClaudeTools } from "../utils/claudeCloaking.js";
 import { filterToOpenAIFormat } from "./formats/openai.js";
 import { normalizeThinkingConfig } from "../services/provider.js";
 import { applyThinking, captureThinking } from "./concerns/thinkingUnified.js";
+import { captureSessionId } from "../utils/sessionManager.js";
 import { AntigravityExecutor } from "../executors/antigravity.js";
 import { PROVIDERS } from "../providers/index.js";
 
@@ -68,6 +69,11 @@ export function translateRequest(sourceFormat, targetFormat, model, body, stream
   // format conversion strips/renames the fields. Applied after translation.
   const thinkingIntent = captureThinking(result);
 
+  // Capture session id from the original body (envelope still intact, e.g. antigravity request.sessionId)
+  const clientSessionId = captureSessionId(result, credentials, connectionId, targetFormat);
+  // Expose to downstream translators (gemini-cli/antigravity envelopes) that run after envelope is stripped
+  if (credentials) credentials._clientSessionId = clientSessionId;
+
   // If same format, skip translation steps
   if (sourceFormat !== targetFormat) {
     // Step 1: source -> openai (if source is not openai)
@@ -101,7 +107,7 @@ export function translateRequest(sourceFormat, targetFormat, model, body, stream
   // Final step: prepare request for Claude format endpoints
   if (targetFormat === FORMATS.CLAUDE) {
     const apiKey = credentials?.accessToken || credentials?.apiKey || null;
-    result = prepareClaudeRequest(result, provider, apiKey, connectionId);
+    result = prepareClaudeRequest(result, provider, apiKey, connectionId, credentials?.rawHeaders, clientSessionId);
   }
 
   // Claude cloaking: rename client tools with _cc suffix (anti-ban)
diff --git a/open-sse/translator/request/openai-to-gemini.js b/open-sse/translator/request/openai-to-gemini.js
index 90db3afe..fe181631 100644
--- a/open-sse/translator/request/openai-to-gemini.js
+++ b/open-sse/translator/request/openai-to-gemini.js
@@ -17,7 +17,7 @@ import {
   generateProjectId,
   cleanJSONSchemaForAntigravity
 } from "../formats/gemini.js";
-import { deriveSessionId } from "../../utils/sessionManager.js";
+import { deriveSessionId, toNumericSessionId } from "../../utils/sessionManager.js";
 import { ROLE, GEMINI_ROLE, OPENAI_BLOCK, CLAUDE_BLOCK } from "../schema/index.js";
 
 // Sanitize function names for Gemini API.
@@ -259,7 +259,7 @@ function wrapInCloudCodeEnvelope(model, geminiCLI, credentials = null, isAntigra
     userAgent: isAntigravity ? "antigravity" : "gemini-cli",
     requestId: isAntigravity ? `agent-${generateUUID()}` : generateRequestId(),
     request: {
-      sessionId: isAntigravity ? deriveSessionId(credentials?.email || credentials?.connectionId) : generateSessionId(),
+      sessionId: toNumericSessionId(credentials?._clientSessionId) || (isAntigravity ? deriveSessionId(credentials?.email || credentials?.connectionId) : generateSessionId()),
       contents: geminiCLI.contents,
       systemInstruction: geminiCLI.systemInstruction,
       generationConfig: geminiCLI.generationConfig,
@@ -309,7 +309,7 @@ function wrapInCloudCodeEnvelopeForClaude(model, claudeRequest, credentials = nu
     requestId: `agent-${generateUUID()}`,
     requestType: "agent",
     request: {
-      sessionId: deriveSessionId(credentials?.email || credentials?.connectionId),
+      sessionId: toNumericSessionId(credentials?._clientSessionId) || deriveSessionId(credentials?.email || credentials?.connectionId),
       contents: [],
       generationConfig: {
         temperature: claudeRequest.temperature || 1,
diff --git a/open-sse/utils/claudeCloaking.js b/open-sse/utils/claudeCloaking.js
index 204a7cfa..46a44e4f 100644
--- a/open-sse/utils/claudeCloaking.js
+++ b/open-sse/utils/claudeCloaking.js
@@ -13,11 +13,18 @@ function generateBillingHeader(payload) {
   return `x-anthropic-billing-header: cc_version=${CLAUDE_VERSION}.${buildHash}; cc_entrypoint=${CC_ENTRYPOINT}; cch=${cch};`;
 }
 
+// Derive a deterministic UUID-v4-shaped string from a seed (stable per account)
+function deriveUuid(seed) {
+  const h = createHash("sha256").update(seed).digest("hex");
+  return `${h.slice(0, 8)}-${h.slice(8, 12)}-4${h.slice(13, 16)}-${((parseInt(h[16], 16) & 0x3) | 0x8).toString(16)}${h.slice(17, 20)}-${h.slice(20, 32)}`;
+}
+
 // Generate fake user ID in Claude Code 2.1.92+ JSON format:
 // {"device_id":"<64hex>","account_uuid":"<uuid>","session_id":"<uuid>"}
-function generateFakeUserID(sessionId) {
-  const deviceId = randomBytes(32).toString("hex");
-  const accountUuid = randomUUID();
+// device_id/account_uuid derive from apiKey (stable per account), session_id per-conversation
+function generateFakeUserID(sessionId, apiKey) {
+  const deviceId = apiKey ? createHash("sha256").update(`device:${apiKey}`).digest("hex") : randomBytes(32).toString("hex");
+  const accountUuid = apiKey ? deriveUuid(`account:${apiKey}`) : randomUUID();
   const sessionUuid = sessionId || randomUUID();
   return `{"device_id":"${deviceId}","account_uuid":"${accountUuid}","session_id":"${sessionUuid}"}`;
 }
@@ -151,7 +158,7 @@ export function applyCloaking(body, apiKey, sessionId) {
   // Inject fake user ID into metadata (session_id must match X-Claude-Code-Session-Id)
   const existingUserId = result.metadata?.user_id;
   if (!existingUserId) {
-    result.metadata = { ...result.metadata, user_id: generateFakeUserID(sessionId) };
+    result.metadata = { ...result.metadata, user_id: generateFakeUserID(sessionId, apiKey) };
   }
 
   return result;
diff --git a/open-sse/utils/sessionManager.js b/open-sse/utils/sessionManager.js
index d8663bf5..73af1a6b 100644
--- a/open-sse/utils/sessionManager.js
+++ b/open-sse/utils/sessionManager.js
@@ -89,7 +89,7 @@ const ASSISTANT_CAP_LEN = 200;
 const MAX_ASSISTANT_SESSIONS = 5000;
 
 // Client headers/body fields that carry an upstream session id (priority order)
-const SESSION_HEADER_KEYS = ["x-session-id", "session_id", "x-amp-thread-id", "x-client-request-id"];
+const SESSION_HEADER_KEYS = ["x-session-id", "session-id", "session_id", "x-amp-thread-id", "x-client-request-id"];
 const CLAUDE_CODE_SESSION_RE = /_session_([a-f0-9-]+)$/;
 
 function sha16(text) {
@@ -122,9 +122,20 @@ function headerValue(headers, key) {
 }
 
 // Read client-provided session id from headers/body (no generation)
+// Antigravity envelope carries session in request.sessionId; requestId embeds conversation uuid
+const ANTIGRAVITY_CONV_RE = /^[a-z]+\/([0-9a-f-]{36})\//i;
+function extractAntigravitySession(body) {
+    const sid = body?.request?.sessionId;
+    if (sid != null && sid !== "") return normalizeSessionId(String(sid));
+    const m = typeof body?.requestId === "string" ? body.requestId.match(ANTIGRAVITY_CONV_RE) : null;
+    return m ? normalizeSessionId(m[1]) : null;
+}
+
 function extractClientSessionId(headers, body) {
     const claude = extractClaudeCodeSession(body?.metadata?.user_id);
     if (claude) return `claude:${claude}`;
+    const antigravity = extractAntigravitySession(body);
+    if (antigravity) return `antigravity:${antigravity}`;
     for (const key of SESSION_HEADER_KEYS) {
         const v = headerValue(headers, key);
         if (v) return v;
@@ -194,6 +205,22 @@ export function resolveSessionId({ headers, body, connectionId, workspaceId, sco
     return deriveSessionId(connectionId);
 }
 
+// Capture session id from request body + credentials (envelope still intact here)
+export function captureSessionId(body, credentials, connectionId, scope = "") {
+    return resolveSessionId({ headers: credentials?.rawHeaders, body, connectionId, scope });
+}
+
+// Convert any session id to Antigravity numeric format "-<int64>" (matches real AG / CLIProxyAPI).
+// Already-numeric ids (native AG sessionId) pass through unchanged.
+export function toNumericSessionId(sessionId) {
+    const v = normalizeSessionId(sessionId);
+    if (!v) return null;
+    if (/^-?\d+$/.test(v)) return v;
+    const h = crypto.createHash("sha256").update(v).digest();
+    const n = h.readBigUInt64BE(0) & 0x7fffffffffffffffn;
+    return `-${n.toString()}`;
+}
+
 // Cleanup expired assistant-session entries
 const assistantCleanup = setInterval(() => {
     const now = Date.now();
diff --git a/src/app/(dashboard)/dashboard/providers/[id]/page.js b/src/app/(dashboard)/dashboard/providers/[id]/page.js
index c06fbaa5..ac1e808b 100644
--- a/src/app/(dashboard)/dashboard/providers/[id]/page.js
+++ b/src/app/(dashboard)/dashboard/providers/[id]/page.js
@@ -48,7 +48,7 @@ export default function ProviderDetailPage() {
   const [headerImgError, setHeaderImgError] = useState(false);
   const [modelTestResults, setModelTestResults] = useState({});
   const [modelsTestError, setModelsTestError] = useState("");
-  const [testingModelId, setTestingModelId] = useState(null);
+  const [testingModelIds, setTestingModelIds] = useState(() => new Set());
   const [showAddCustomModel, setShowAddCustomModel] = useState(false);
   const [selectedConnectionIds, setSelectedConnectionIds] = useState([]);
   const [bulkProxyPoolId, setBulkProxyPoolId] = useState("__none__");
@@ -876,8 +876,8 @@ export default function ProviderDetailPage() {
   );
 
   const handleTestModel = async (modelId) => {
-    if (testingModelId) return;
-    setTestingModelId(modelId);
+    if (testingModelIds.has(modelId)) return;
+    setTestingModelIds((prev) => new Set(prev).add(modelId));
     try {
       const res = await fetch("/api/models/test", {
         method: "POST",
@@ -891,7 +891,7 @@ export default function ProviderDetailPage() {
       setModelTestResults((prev) => ({ ...prev, [modelId]: "error" }));
       setModelsTestError("Network error");
     } finally {
-      setTestingModelId(null);
+      setTestingModelIds((prev) => { const n = new Set(prev); n.delete(modelId); return n; });
     }
   };
 
@@ -952,7 +952,7 @@ export default function ProviderDetailPage() {
             onDeleteAlias={() => handleDeleteAlias(model.alias)}
             testStatus={modelTestResults[model.id]}
             onTest={connections.length > 0 || isFreeNoAuth ? () => handleTestModel(model.id) : undefined}
-            isTesting={testingModelId === model.id}
+            isTesting={testingModelIds.has(model.id)}
             isCustom
             isFree={false}
             caps={getCaps(`${providerId}/${model.id}`)}
@@ -977,7 +977,7 @@ export default function ProviderDetailPage() {
               onDeleteAlias={() => handleDeleteAlias(existingAlias)}
               testStatus={modelTestResults[model.id]}
               onTest={connections.length > 0 || isFreeNoAuth ? () => handleTestModel(model.id) : undefined}
-              isTesting={testingModelId === model.id}
+              isTesting={testingModelIds.has(model.id)}
               isFree={model.isFree}
               onDisable={() => handleDisableModel(model.id)}
               caps={getCaps(`${providerId}/${model.id}`)}
diff --git a/src/app/(dashboard)/dashboard/providers/page.js b/src/app/(dashboard)/dashboard/providers/page.js
index 53581ccc..e911ac90 100644
--- a/src/app/(dashboard)/dashboard/providers/page.js
+++ b/src/app/(dashboard)/dashboard/providers/page.js
@@ -281,15 +281,18 @@ export default function ProvidersPage() {
     Object.entries(OAUTH_PROVIDERS).filter(([, info]) => !info.hidden && matchSearch(info.name)),
     "oauth",
   );
-  const freeEntries = Object.entries(FREE_PROVIDERS).filter(
-    ([, info]) => !info.hidden && matchSearch(info.name),
-  );
+  const freeEntries = Object.entries(FREE_PROVIDERS)
+    .filter(([, info]) => !info.hidden && matchSearch(info.name))
+    .sort(([, a], [, b]) => (b.noAuth ? 1 : 0) - (a.noAuth ? 1 : 0));
   const freeTierEntries = sortByPriority(
     Object.entries(FREE_TIER_PROVIDERS).filter(
-      ([, info]) => !info.hidden && matchSearch(info.name),
+      ([, info]) =>
+        !info.hidden &&
+        matchSearch(info.name) &&
+        (info.serviceKinds ?? ["llm"]).includes("llm"),
     ),
     "freeTier",
-  );
+  ).sort(([, a], [, b]) => (b.noAuth ? 1 : 0) - (a.noAuth ? 1 : 0));
   // API Key: connected providers first, then alphabetical by name
   const apikeyEntries = Object.entries(APIKEY_PROVIDERS)
     .filter(
diff --git a/src/shared/hooks/useModelCaps.js b/src/shared/hooks/useModelCaps.js
index db16bdda..3dbdcdd4 100644
--- a/src/shared/hooks/useModelCaps.js
+++ b/src/shared/hooks/useModelCaps.js
@@ -1,6 +1,7 @@
 "use client";
 
 import { useState, useEffect } from "react";
+import { getCapabilitiesForModel } from "open-sse/providers/capabilities.js";
 
 // Fetch model capabilities once and expose a lookup by fullModel ("provider/model") or bare model id.
 export function useModelCaps() {
@@ -32,7 +33,11 @@ export function useModelCaps() {
     if (!key) return null;
     if (byFull[key]) return byFull[key];
     const bare = key.includes("/") ? key.slice(key.indexOf("/") + 1) : key;
-    return byId[bare] || null;
+    if (byId[bare]) return byId[bare];
+    // Fallback: compute caps for dynamic models (passthrough/custom/suggested) not in static list
+    const provider = key.includes("/") ? key.slice(0, key.indexOf("/")) : null;
+    const c = getCapabilitiesForModel(provider, bare);
+    return { vision: c.vision, search: c.search, reasoning: c.reasoning };
   };
 
   return { getCaps };