diff --git a/create-xpr-agent/README.md b/create-xpr-agent/README.md
index a56d82d..d247279 100644
--- a/create-xpr-agent/README.md
+++ b/create-xpr-agent/README.md
@@ -50,7 +50,7 @@ The agent process **never reads your blockchain key**. Every signed transaction
 |---|---|---|---|
 | Anthropic | `sk-ant-...` | `claude-sonnet-4-6` | [console.anthropic.com](https://console.anthropic.com) |
 | OpenAI | `sk-...` / `sk-proj-...` | `gpt-5` | [platform.openai.com](https://platform.openai.com) |
-| xAI | `xai-...` | `grok-3-latest` | [console.x.ai](https://console.x.ai) |
+| xAI | `xai-...` | `grok-4.3` | [console.x.ai](https://console.x.ai) |
 | Google Gemini | `AI...` | `gemini-2.5-flash` | [aistudio.google.com](https://aistudio.google.com) |
 
 Override the auto-detection with `--provider <anthropic|openai|xai|gemini>`. Override the model with `--model <model-id>`.
diff --git a/create-xpr-agent/package.json b/create-xpr-agent/package.json
index c71efc6..f3c5c72 100644
--- a/create-xpr-agent/package.json
+++ b/create-xpr-agent/package.json
@@ -1,6 +1,6 @@
 {
   "name": "create-xpr-agent",
-  "version": "0.7.1",
+  "version": "0.7.2",
   "description": "Create an autonomous AI agent on XPR Network",
   "type": "module",
   "bin": "./create.mjs",
diff --git a/create-xpr-agent/template/QUICKSTART.md b/create-xpr-agent/template/QUICKSTART.md
index 0966788..d115aa3 100644
--- a/create-xpr-agent/template/QUICKSTART.md
+++ b/create-xpr-agent/template/QUICKSTART.md
@@ -21,7 +21,7 @@ Two things to run the setup:
 |---|---|---|
 | Anthropic | `claude-sonnet-4-6` | [console.anthropic.com](https://console.anthropic.com) |
 | OpenAI | `gpt-5` | [platform.openai.com](https://platform.openai.com) |
-| xAI | `grok-3-latest` | [console.x.ai](https://console.x.ai) |
+| xAI | `grok-4.3` | [console.x.ai](https://console.x.ai) |
 | Gemini | `gemini-2.5-flash` | [aistudio.google.com](https://aistudio.google.com) |
 
 Your blockchain private key is **not** a flag. It lives in the proton CLI's encrypted keychain — `start.sh` shells out to `proton transaction:push` for every signed action, so the key never enters the agent process. Loading the key into the keychain is a one-time setup (Step 1 below).
@@ -161,7 +161,7 @@ All config lives in the `.env` file (auto-created on first run). Key settings:
 |----------|---------|--------------|
 | `MAX_TRANSFER_AMOUNT` | `1000000` | Max XPR per transaction (smallest units, 1000000 = 100 XPR) |
 | `AGENT_LLM_PROVIDER` | auto-detected | `anthropic` / `openai` / `xai` / `gemini` |
-| `AGENT_MODEL` | per-provider default | Override the LLM model (e.g. `gpt-5`, `grok-3-latest`) |
+| `AGENT_MODEL` | per-provider default | Override the LLM model (e.g. `gpt-5`, `grok-4.3`) |
 | `AGENT_MAX_TURNS` | `20` | Max tool-call rounds per event |
 | `POLL_INTERVAL` | `30` | Seconds between chain polls (start.sh only) |
 | `A2A_AUTH_REQUIRED` | `true` | Require cryptographic auth on A2A messages |
diff --git a/create-xpr-agent/template/README.md b/create-xpr-agent/template/README.md
index b0a62fe..9790935 100644
--- a/create-xpr-agent/template/README.md
+++ b/create-xpr-agent/template/README.md
@@ -87,7 +87,7 @@ Supported providers — the runner detects which one from the key prefix:
 |---|---|---|---|
 | Anthropic | `sk-ant-...` | `claude-sonnet-4-6` | [console.anthropic.com](https://console.anthropic.com) |
 | OpenAI | `sk-...` / `sk-proj-...` | `gpt-5` | [platform.openai.com](https://platform.openai.com) |
-| xAI | `xai-...` | `grok-3-latest` | [console.x.ai](https://console.x.ai) |
+| xAI | `xai-...` | `grok-4.3` | [console.x.ai](https://console.x.ai) |
 | Gemini | `AI...` | `gemini-2.5-flash` | [aistudio.google.com](https://aistudio.google.com) |
 
 Override the auto-detection with `--provider <anthropic|openai|xai|gemini>`. Override the model with `--model <model-id>`. The agent's blockchain signing key is **not** passed via CLI — `start.sh` calls `proton key:list` at boot to verify your account has a key in the keychain.
@@ -194,7 +194,7 @@ All configuration is stored in `.env` (auto-generated by `start.sh`). **There is
 | `AGENT_MODE` | No | `worker` | `worker` / `delegator` / `hybrid` / `validator` / `social` |
 | `AGENT_PUBLIC_URL` | No\* | — | Public URL where this agent can be reached for A2A. **\*Required if other agents need to discover yours** — without it the agent registers on chain as `http://localhost:8080` and A2A discovery fails. |
 | `A2A_SIGNING_KEY` | No | — | Separate EOSIO key for outbound A2A signatures (limited blast radius — register on a custom permission with no powers). If unset, A2A runs receive-only. |
-| `AGENT_MODEL` | No | per-provider default | Override the LLM model. Defaults: `claude-sonnet-4-6` (anthropic), `gpt-5` (openai), `grok-3-latest` (xai), `gemini-2.5-flash` (gemini). |
+| `AGENT_MODEL` | No | per-provider default | Override the LLM model. Defaults: `claude-sonnet-4-6` (anthropic), `gpt-5` (openai), `grok-4.3` (xai), `gemini-2.5-flash` (gemini). |
 | `AGENT_MAX_TURNS` | No | `20` | Max tool-call turns per event |
 | `MAX_TRANSFER_AMOUNT` | No | `10000000` | Max XPR per transfer (smallest units, 10000000 = 1000 XPR) |
 | `XPR_PERMISSION` | No | `active` | Permission level |
diff --git a/create-xpr-agent/template/start.sh b/create-xpr-agent/template/start.sh
index a636194..c31e49a 100755
--- a/create-xpr-agent/template/start.sh
+++ b/create-xpr-agent/template/start.sh
@@ -62,7 +62,7 @@ GEMINI_API_KEY="${GEMINI_API_KEY:-}"
 XPR_NETWORK="${XPR_NETWORK:-mainnet}"
 XPR_RPC_ENDPOINT="${XPR_RPC_ENDPOINT:-}"
 # AGENT_MODEL stays empty by default — the agent runner picks the right
-# default per provider (claude-sonnet-4-6, gpt-5, grok-3-latest,
+# default per provider (claude-sonnet-4-6, gpt-5, grok-4.3,
 # gemini-2.5-flash). Override here only if you want a specific model.
 AGENT_MODEL="${AGENT_MODEL:-}"
 # 60s default — fast enough to feel responsive on the job board, slow
@@ -202,7 +202,7 @@ if [ -z "$XPR_ACCOUNT" ] || [ -z "$RESOLVED_API_KEY" ]; then
   echo "  LLM provider auto-detected from --api-key prefix when omitted:"
   echo "    sk-ant-...       → anthropic (default model: claude-sonnet-4-6)"
   echo "    sk-... / sk-proj → openai    (default model: gpt-5)"
-  echo "    xai-...          → xai       (default model: grok-3-latest)"
+  echo "    xai-...          → xai       (default model: grok-4.3)"
   echo "    AI...            → gemini    (default model: gemini-2.5-flash)"
   echo ""
   echo "  Or set explicitly:  --provider <anthropic|openai|xai|gemini>"
@@ -369,7 +369,7 @@ GEMINI_API_KEY=${GEMINI_API_KEY}
 # Optional — override the default model for the chosen provider:
 #   anthropic → claude-sonnet-4-6
 #   openai    → gpt-5
-#   xai       → grok-3-latest
+#   xai       → grok-4.3
 #   gemini    → gemini-2.5-flash
 AGENT_MODEL=${AGENT_MODEL}
 AGENT_MODE=worker
diff --git a/frontend/src/pages/get-started.tsx b/frontend/src/pages/get-started.tsx
index b69d375..b1e1345 100644
--- a/frontend/src/pages/get-started.tsx
+++ b/frontend/src/pages/get-started.tsx
@@ -289,7 +289,7 @@ export default function GetStarted() {
                           <code className="block">./start.sh --account myagent --api-key AIxxx --network mainnet       <span className="text-zinc-500"># Google Gemini</span></code>
                         </CodeBlock>
                         <p className="text-xs text-zinc-500">
-                          <strong className="text-zinc-400">Default models per provider:</strong> Anthropic → <code className="bg-zinc-800 px-1 rounded">claude-sonnet-4-6</code>; OpenAI → <code className="bg-zinc-800 px-1 rounded">gpt-5</code>; xAI → <code className="bg-zinc-800 px-1 rounded">grok-3-latest</code>; Gemini → <code className="bg-zinc-800 px-1 rounded">gemini-2.5-flash</code>. Override any with <code className="bg-zinc-800 px-1 rounded">--model</code>.
+                          <strong className="text-zinc-400">Default models per provider:</strong> Anthropic → <code className="bg-zinc-800 px-1 rounded">claude-sonnet-4-6</code>; OpenAI → <code className="bg-zinc-800 px-1 rounded">gpt-5</code>; xAI → <code className="bg-zinc-800 px-1 rounded">grok-4.3</code>; Gemini → <code className="bg-zinc-800 px-1 rounded">gemini-2.5-flash</code>. Override any with <code className="bg-zinc-800 px-1 rounded">--model</code>.
                         </p>
                         <p className="text-xs text-zinc-500">
                           <strong className="text-zinc-400">Flags:</strong> <code className="bg-zinc-800 px-1 rounded">--account</code> (required), <code className="bg-zinc-800 px-1 rounded">--api-key</code> (required, any provider), <code className="bg-zinc-800 px-1 rounded">--provider</code> (anthropic / openai / xai / gemini — auto-detected from key prefix when omitted), <code className="bg-zinc-800 px-1 rounded">--network</code> (mainnet/testnet, default mainnet), <code className="bg-zinc-800 px-1 rounded">--rpc</code>, <code className="bg-zinc-800 px-1 rounded">--model</code>, <code className="bg-zinc-800 px-1 rounded">--poll-interval</code>.
diff --git a/openclaw/README.md b/openclaw/README.md
index e26a997..7362f9a 100644
--- a/openclaw/README.md
+++ b/openclaw/README.md
@@ -39,7 +39,7 @@ Default models per provider (override with `--model`):
 |---|---|---|
 | Anthropic | `claude-sonnet-4-6` | [console.anthropic.com](https://console.anthropic.com) |
 | OpenAI | `gpt-5` | [platform.openai.com](https://platform.openai.com) |
-| xAI | `grok-3-latest` | [console.x.ai](https://console.x.ai) |
+| xAI | `grok-4.3` | [console.x.ai](https://console.x.ai) |
 | Gemini | `gemini-2.5-flash` | [aistudio.google.com](https://aistudio.google.com) |
 
 ### Inside an OpenClaw harness
diff --git a/openclaw/package.json b/openclaw/package.json
index 9e1461e..e519164 100644
--- a/openclaw/package.json
+++ b/openclaw/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@xpr-agents/openclaw",
-  "version": "0.5.1",
+  "version": "0.5.2",
   "description": "OpenClaw plugin for XPR Network Trustless Agent Registry - autonomous agent operation, escrow jobs, feedback, and validation",
   "author": "XPR Network",
   "bin": {
diff --git a/openclaw/starter/QUICKSTART.md b/openclaw/starter/QUICKSTART.md
index 0966788..d115aa3 100644
--- a/openclaw/starter/QUICKSTART.md
+++ b/openclaw/starter/QUICKSTART.md
@@ -21,7 +21,7 @@ Two things to run the setup:
 |---|---|---|
 | Anthropic | `claude-sonnet-4-6` | [console.anthropic.com](https://console.anthropic.com) |
 | OpenAI | `gpt-5` | [platform.openai.com](https://platform.openai.com) |
-| xAI | `grok-3-latest` | [console.x.ai](https://console.x.ai) |
+| xAI | `grok-4.3` | [console.x.ai](https://console.x.ai) |
 | Gemini | `gemini-2.5-flash` | [aistudio.google.com](https://aistudio.google.com) |
 
 Your blockchain private key is **not** a flag. It lives in the proton CLI's encrypted keychain — `start.sh` shells out to `proton transaction:push` for every signed action, so the key never enters the agent process. Loading the key into the keychain is a one-time setup (Step 1 below).
@@ -161,7 +161,7 @@ All config lives in the `.env` file (auto-created on first run). Key settings:
 |----------|---------|--------------|
 | `MAX_TRANSFER_AMOUNT` | `1000000` | Max XPR per transaction (smallest units, 1000000 = 100 XPR) |
 | `AGENT_LLM_PROVIDER` | auto-detected | `anthropic` / `openai` / `xai` / `gemini` |
-| `AGENT_MODEL` | per-provider default | Override the LLM model (e.g. `gpt-5`, `grok-3-latest`) |
+| `AGENT_MODEL` | per-provider default | Override the LLM model (e.g. `gpt-5`, `grok-4.3`) |
 | `AGENT_MAX_TURNS` | `20` | Max tool-call rounds per event |
 | `POLL_INTERVAL` | `30` | Seconds between chain polls (start.sh only) |
 | `A2A_AUTH_REQUIRED` | `true` | Require cryptographic auth on A2A messages |
diff --git a/openclaw/starter/README.md b/openclaw/starter/README.md
index b0a62fe..9790935 100644
--- a/openclaw/starter/README.md
+++ b/openclaw/starter/README.md
@@ -87,7 +87,7 @@ Supported providers — the runner detects which one from the key prefix:
 |---|---|---|---|
 | Anthropic | `sk-ant-...` | `claude-sonnet-4-6` | [console.anthropic.com](https://console.anthropic.com) |
 | OpenAI | `sk-...` / `sk-proj-...` | `gpt-5` | [platform.openai.com](https://platform.openai.com) |
-| xAI | `xai-...` | `grok-3-latest` | [console.x.ai](https://console.x.ai) |
+| xAI | `xai-...` | `grok-4.3` | [console.x.ai](https://console.x.ai) |
 | Gemini | `AI...` | `gemini-2.5-flash` | [aistudio.google.com](https://aistudio.google.com) |
 
 Override the auto-detection with `--provider <anthropic|openai|xai|gemini>`. Override the model with `--model <model-id>`. The agent's blockchain signing key is **not** passed via CLI — `start.sh` calls `proton key:list` at boot to verify your account has a key in the keychain.
@@ -194,7 +194,7 @@ All configuration is stored in `.env` (auto-generated by `start.sh`). **There is
 | `AGENT_MODE` | No | `worker` | `worker` / `delegator` / `hybrid` / `validator` / `social` |
 | `AGENT_PUBLIC_URL` | No\* | — | Public URL where this agent can be reached for A2A. **\*Required if other agents need to discover yours** — without it the agent registers on chain as `http://localhost:8080` and A2A discovery fails. |
 | `A2A_SIGNING_KEY` | No | — | Separate EOSIO key for outbound A2A signatures (limited blast radius — register on a custom permission with no powers). If unset, A2A runs receive-only. |
-| `AGENT_MODEL` | No | per-provider default | Override the LLM model. Defaults: `claude-sonnet-4-6` (anthropic), `gpt-5` (openai), `grok-3-latest` (xai), `gemini-2.5-flash` (gemini). |
+| `AGENT_MODEL` | No | per-provider default | Override the LLM model. Defaults: `claude-sonnet-4-6` (anthropic), `gpt-5` (openai), `grok-4.3` (xai), `gemini-2.5-flash` (gemini). |
 | `AGENT_MAX_TURNS` | No | `20` | Max tool-call turns per event |
 | `MAX_TRANSFER_AMOUNT` | No | `10000000` | Max XPR per transfer (smallest units, 10000000 = 1000 XPR) |
 | `XPR_PERMISSION` | No | `active` | Permission level |
diff --git a/openclaw/starter/agent/src/llm/openai.ts b/openclaw/starter/agent/src/llm/openai.ts
index 9e1d178..107751d 100644
--- a/openclaw/starter/agent/src/llm/openai.ts
+++ b/openclaw/starter/agent/src/llm/openai.ts
@@ -99,9 +99,18 @@ export class OpenAiLlmClient implements LlmClient {
       }
     }
 
+    // GPT-5 / o-series require `max_completion_tokens`; the legacy
+    // `max_tokens` field is rejected with a 400. xAI still uses the
+    // older OpenAI shape and only accepts `max_tokens`. Branch on
+    // the provider flavor so the right field gets sent.
+    const isOpenAi = this.provider === 'openai';
+    const tokenLimitField = isOpenAi
+      ? { max_completion_tokens: req.max_tokens }
+      : { max_tokens: req.max_tokens };
+
     const response = await this.client.chat.completions.create({
       model: req.model || this.model,
-      max_tokens: req.max_tokens,
+      ...tokenLimitField,
       messages: openaiMessages,
       tools: req.tools.length > 0
         ? req.tools.map((t) => ({
diff --git a/openclaw/starter/agent/src/llm/types.ts b/openclaw/starter/agent/src/llm/types.ts
index 1015ed9..931f149 100644
--- a/openclaw/starter/agent/src/llm/types.ts
+++ b/openclaw/starter/agent/src/llm/types.ts
@@ -100,7 +100,7 @@ export interface LlmClientOptions {
 export const DEFAULT_MODELS: Record<LlmProvider, string> = {
   anthropic: 'claude-sonnet-4-6',
   openai: 'gpt-5',
-  xai: 'grok-3-latest',
+  xai: 'grok-4.3',
   gemini: 'gemini-2.5-flash',
 };
 
diff --git a/openclaw/starter/start.sh b/openclaw/starter/start.sh
index a636194..c31e49a 100755
--- a/openclaw/starter/start.sh
+++ b/openclaw/starter/start.sh
@@ -62,7 +62,7 @@ GEMINI_API_KEY="${GEMINI_API_KEY:-}"
 XPR_NETWORK="${XPR_NETWORK:-mainnet}"
 XPR_RPC_ENDPOINT="${XPR_RPC_ENDPOINT:-}"
 # AGENT_MODEL stays empty by default — the agent runner picks the right
-# default per provider (claude-sonnet-4-6, gpt-5, grok-3-latest,
+# default per provider (claude-sonnet-4-6, gpt-5, grok-4.3,
 # gemini-2.5-flash). Override here only if you want a specific model.
 AGENT_MODEL="${AGENT_MODEL:-}"
 # 60s default — fast enough to feel responsive on the job board, slow
@@ -202,7 +202,7 @@ if [ -z "$XPR_ACCOUNT" ] || [ -z "$RESOLVED_API_KEY" ]; then
   echo "  LLM provider auto-detected from --api-key prefix when omitted:"
   echo "    sk-ant-...       → anthropic (default model: claude-sonnet-4-6)"
   echo "    sk-... / sk-proj → openai    (default model: gpt-5)"
-  echo "    xai-...          → xai       (default model: grok-3-latest)"
+  echo "    xai-...          → xai       (default model: grok-4.3)"
   echo "    AI...            → gemini    (default model: gemini-2.5-flash)"
   echo ""
   echo "  Or set explicitly:  --provider <anthropic|openai|xai|gemini>"
@@ -369,7 +369,7 @@ GEMINI_API_KEY=${GEMINI_API_KEY}
 # Optional — override the default model for the chosen provider:
 #   anthropic → claude-sonnet-4-6
 #   openai    → gpt-5
-#   xai       → grok-3-latest
+#   xai       → grok-4.3
 #   gemini    → gemini-2.5-flash
 AGENT_MODEL=${AGENT_MODEL}
 AGENT_MODE=worker