Grok 4.20 Multi-Agent·续

Grok 4.20 Multi-Agent·续

xAI released major voice and audio updates, including Custom Voices/Voice Library (allowing voice cloning from short recordings), Grok Voice Think Fast 1.0 (a capable voice agent for API use), and Speech-to-Text/Text-to-Speech APIs with natural voices, multilingual support, and simple pricing (announced April 23–30).[2]
Grok 4.3 launched with strong benchmarks and aggressive pricing, scoring 53 on the Artificial Analysis Intelligence Index (improved agentic/tool-use performance, ~4 points ahead of its predecessor), alongside ~40–60% price cuts; this ties into broader product momentum and the post-acquisition context, though some reports noted employee departures and prior coding lags.[4]

These reflect xAI’s rapid product iteration in voice/agent capabilities and model performance while navigating legal, integration, and operational developments following the SpaceX deal.


```json
{
  "background": false,
  "completed_at": 1777680547,
  "created_at": 0,
  "error": null,
  "frequency_penalty": 0,
  "gatewayMetadata": {
    "keySource": "Unified"
  },
  "id": "ee8dcd35-f5a1-9a07-924c-db90a88da556",
  "incomplete_details": null,
  "instructions": null,
  "max_output_tokens": null,
  "max_tool_calls": null,
  "metadata": {
    "system_fingerprint": ""
  },
  "model": "grok-4.20-multi-agent-0309",
  "object": "response",
  "output": [
    {
      "action": {
        "query": "xAI news April 2026 OR May 2026",
        "sources": [],
        "type": "search"
      },
      "id": "ws_ee8dcd35-f5a1-9a07-924c-db90a88da556_0",
      "status": "completed",
      "type": "web_search_call"
    },
    {
      "action": {
        "query": "xAI Grok latest news this week",
        "sources": [],
        "type": "search"
      },
      "id": "ws_ee8dcd35-f5a1-9a07-924c-db90a88da556_1",
      "status": "completed",
      "type": "web_search_call"
    },
    {
      "action": {
        "query": "\"xAI\" (announcement OR launch OR funding OR Grok) after:2026-04-20",
        "sources": [],
        "type": "search"
      },
      "id": "ws_ee8dcd35-f5a1-9a07-924c-db90a88da556_2",
      "status": "completed",
      "type": "web_search_call"
    },
    {
      "id": "tco_ee8dcd35-f5a1-9a07-924c-db90a88da556",
      "status": "completed",
      "summary": [],
      "type": "reasoning"
    },
    {
      "id": "tco_ee8dcd35-f5a1-9a07-924c-db90a88da556",
      "status": "completed",
      "summary": [],
      "type": "reasoning"
    },
    {
      "id": "tco_ee8dcd35-f5a1-9a07-924c-db90a88da556",
      "status": "completed",
      "summary": [],
      "type": "reasoning"
    },
    {
      "action": {
        "type": "open_page",
        "url": "https://x.ai/news"
      },
      "id": "ws_ee8dcd35-f5a1-9a07-924c-db90a88da556_3",
      "status": "completed",
      "type": "web_search_call"
    },
    {
      "action": {
        "query": "Elon Musk testifies xAI OpenAI Grok April 2026 OR May 2026",
        "sources": [],
        "type": "search"
      },
      "id": "ws_ee8dcd35-f5a1-9a07-924c-db90a88da556_4",
      "status": "completed",
      "type": "web_search_call"
    },
    {
      "id": "tco_ee8dcd35-f5a1-9a07-924c-db90a88da556",
      "status": "completed",
      "summary": [],
      "type": "reasoning"
    },
    {
      "id": "tco_ee8dcd35-f5a1-9a07-924c-db90a88da556",
      "status": "completed",
      "summary": [],
      "type": "reasoning"
    },
    {
      "content": [
        {
          "annotations": [
            {
              "end_index": 166,
              "start_index": 66,
              "title": "1",
              "type": "url_citation",
              "url": "https://techcrunch.com/2026/04/30/elon-musk-testifies-that-xai-trained-grok-on-openai-models/"
            },
            {
              "end_index": 190,
              "start_index": 166,
              "title": "2",
              "type": "url_citation",
              "url": "https://x.ai/news"
            },
            {
              "end_index": 617,
              "start_index": 517,
              "title": "1",
              "type": "url_citation",
              "url": "https://techcrunch.com/2026/04/30/elon-musk-testifies-that-xai-trained-grok-on-openai-models/"
            },
            {
              "end_index": 728,
              "start_index": 617,
              "title": "3",
              "type": "url_citation",
              "url": "https://www.theverge.com/ai-artificial-intelligence/921546/elon-musk-xai-openai-trial-model-distillation"
            },
            {
              "end_index": 1078,
              "start_index": 1054,
              "title": "2",
              "type": "url_citation",
              "url": "https://x.ai/news"
            },
            {
              "end_index": 1593,
              "start_index": 1457,
              "title": "4",
              "type": "url_citation",
              "url": "https://venturebeat.com/technology/xai-launches-grok-4-3-at-an-aggressively-low-price-and-a-new-fast-powerful-voice-cloning-suite"
            }
          ],
          "logprobs": [],
          "text": "**Top xAI news stories this week (late April to early May 2026):**[[1]](https://techcrunch.com/2026/04/30/elon-musk-testifies-that-xai-trained-grok-on-openai-models/)[[2]](https://x.ai/news)\n\n- **Elon Musk testified in his lawsuit against OpenAI**, admitting that xAI had “partly” used OpenAI models (via distillation techniques) to train Grok. The multi-day testimony framed Musk as an AI safety advocate contrasting with OpenAI’s for-profit shift; it occurred as xAI operates under its recent acquisition by SpaceX.[[1]](https://techcrunch.com/2026/04/30/elon-musk-testifies-that-xai-trained-grok-on-openai-models/)[[3]](https://www.theverge.com/ai-artificial-intelligence/921546/elon-musk-xai-openai-trial-model-distillation)\n\n- **xAI released major voice and audio updates**, including Custom Voices/Voice Library (allowing voice cloning from short recordings), Grok Voice Think Fast 1.0 (a capable voice agent for API use), and Speech-to-Text/Text-to-Speech APIs with natural voices, multilingual support, and simple pricing (announced April 23–30).[[2]](https://x.ai/news)\n\n- **Grok 4.3 launched with strong benchmarks and aggressive pricing**, scoring 53 on the Artificial Analysis Intelligence Index (improved agentic/tool-use performance, ~4 points ahead of its predecessor), alongside ~40–60% price cuts; this ties into broader product momentum and the post-acquisition context, though some reports noted employee departures and prior coding lags.[[4]](https://venturebeat.com/technology/xai-launches-grok-4-3-at-an-aggressively-low-price-and-a-new-fast-powerful-voice-cloning-suite)\n\nThese reflect xAI’s rapid product iteration in voice/agent capabilities and model performance while navigating legal, integration, and operational developments following the SpaceX deal.",
          "type": "output_text"
        }
      ],
      "id": "msg_ee8dcd35-f5a1-9a07-924c-db90a88da556",
      "role": "assistant",
      "status": "completed",
      "type": "message"
    }
  ],
  "parallel_tool_calls": true,
  "presence_penalty": 0,
  "previous_response_id": null,
  "prompt_cache_key": null,
  "reasoning": {
    "effort": "medium",
    "summary": "detailed"
  },
  "safety_identifier": null,
  "service_tier": "default",
  "status": "completed",
  "store": true,
  "temperature": 0.7,
  "text": {
    "format": {
      "type": "text"
    }
  },
  "tool_choice": "auto",
  "tools": [
    {
      "search_context_size": "medium",
      "type": "web_search"
    }
  ],
  "top_logprobs": 0,
  "top_p": 0.95,
  "truncation": "disabled",
  "usage": {
    "cost_in_usd_ticks": 2026981500,
    "input_tokens": 98165,
    "input_tokens_details": {
      "cached_tokens": 43072
    },
    "num_server_side_tools_used": 21,
    "num_sources_used": 0,
    "output_tokens": 8087,
    "output_tokens_details": {
      "reasoning_tokens": 7740
    },
    "server_side_tool_usage_details": {
      "code_interpreter_calls": 0,
      "document_search_calls": 0,
      "file_search_calls": 0,
      "mcp_calls": 0,
      "web_search_calls": 21,
      "x_search_calls": 0
    },
    "total_tokens": 106252
  },
  "user": null
}

**With Reasoning** — Asking the agent to think harder before responding

const response = await env.AI.run(
  'xai/grok-4.20-multi-agent-0309',
  {
    input:
      'Solve this step by step: Two trains 900 miles apart travel toward each other at 60 and 80 mph. When do they meet?',
    max_output_tokens: 8192,
    reasoning: { effort: 'medium' },
  },
)
console.log(response)

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/v1/responses \
  --header "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "xai/grok-4.20-multi-agent-0309",
  "input": "Solve this step by step: Two trains 900 miles apart travel toward each other at 60 and 80 mph. When do they meet?",
  "max_output_tokens": 8192,
  "reasoning": {
    "effort": "medium"
  }
}'

The two trains are moving toward each other, so the distance between them closes at a combined rate of 60 + 80 = 140 mph.

Time to meet = total distance / combined speed = 900 / 140.

This simplifies (by dividing numerator and denominator by 20) to exactly 45/7 hours (or 6 3/7 hours).

To verify, in that time the first train travels 60 * (45/7) = 2700/7 ≈ 385.71 miles, and the second travels 80 * (45/7) = 3600/7 ≈ 514.29 miles. These add up to exactly 900 miles, confirming they meet after that interval (assuming both start at the same time).

**Final Answer**

45/7 hours

{
  "background": false,
  "completed_at": 1777679680,
  "created_at": 0,
  "error": null,
  "frequency_penalty": 0,
  "gatewayMetadata": {
    "keySource": "Unified"
  },
  "id": "09617b62-0046-cc3e-6d7d-4231238686fb",
  "incomplete_details": null,
  "instructions": null,
  "max_output_tokens": 8192,
  "max_tool_calls": null,
  "metadata": {
    "system_fingerprint": ""
  },
  "model": "grok-4.20-multi-agent-0309",
  "object": "response",
  "output": [
    {
      "content": [
        {
          "annotations": [],
          "logprobs": [],
          "text": "The two trains are moving toward each other, so the distance between them closes at a combined rate of 60 + 80 = 140 mph.\n\nTime to meet = total distance / combined speed = 900 / 140.\n\nThis simplifies (by dividing numerator and denominator by 20) to exactly 45/7 hours (or 6 3/7 hours).\n\nTo verify, in that time the first train travels 60 * (45/7) = 2700/7 ≈ 385.71 miles, and the second travels 80 * (45/7) = 3600/7 ≈ 514.29 miles. These add up to exactly 900 miles, confirming they meet after that interval (assuming both start at the same time).\n\n**Final Answer**\n\n45/7 hours",
          "type": "output_text"
        }
      ],
      "id": "msg_09617b62-0046-cc3e-6d7d-4231238686fb",
      "role": "assistant",
      "status": "completed",
      "type": "message"
    }
  ],
  "parallel_tool_calls": true,
  "presence_penalty": 0,
  "previous_response_id": null,
  "prompt_cache_key": null,
  "reasoning": {
    "effort": "medium",
    "summary": "detailed"
  },
  "safety_identifier": null,
  "service_tier": "default",
  "status": "completed",
  "store": true,
  "temperature": 0.7,
  "text": {
    "format": {
      "type": "text"
    }
  },
  "tool_choice": "auto",
  "tools": [],
  "top_logprobs": 0,
  "top_p": 0.95,
  "truncation": "disabled",
  "usage": {
    "cost_in_usd_ticks": 119112500,
    "input_tokens": 12473,
    "input_tokens_details": {
      "cached_tokens": 9600
    },
    "num_server_side_tools_used": 0,
    "num_sources_used": 0,
    "output_tokens": 2560,
    "output_tokens_details": {
      "reasoning_tokens": 2384
    },
    "total_tokens": 15033
  },
  "user": null
}

Parameters

▶input

one ofrequired

instructions

string

max_output_tokens

integer | null

max_turns

integer | null

parallel_tool_calls

boolean

previous_response_id

string

prompt_cache_key

string

▶reasoning{}

object

▶search_parameters{}

object

store

boolean

stream

boolean

temperature

number | null

▶text{}

object

▶tool_choice

one of

tools

array | null

top_logprobs

integer | null

logprobs

boolean

top_p

number | null

user

string

context_management

array | null

include

array | null

string

object

stringconst: response

created_at

number

completed_at

number

model

string

status

stringenum: in_progress, completed, incomplete, failed

▶output[]

array

parallel_tool_calls

boolean

previous_response_id

string

reasoning

text

tool_choice

▶tools[]

array

top_p

number

temperature

number

instructions

string

incomplete_details

metadata

store

boolean

user

string

error

▶usage{}

object

API Schemas (Raw)

Input Open Download

Output Open Download

Was this helpful?

YesNo

{"@context":"https://schema.org","@type":"TechArticle","@id":"https://developers.cloudflare.com/ai/models/xai/grok-4.20-multi-agent-0309/#page","headline":"Grok 4.20 Multi-Agent","description":"xAI's Grok 4.20 multi-agent model with a 2M-token context window. Multiple agents collaborate in parallel to perform deep research tasks, with function calling, structured outputs, and reasoning capabilities.","url":"https://developers.cloudflare.com/ai/models/xai/grok-4.20-multi-agent-0309/","inLanguage":"en","image":"https://developers.cloudflare.com/og-docs.png","publisher":{"@type":"Organization","name":"Cloudflare","description":"One platform for your apps, agents, and workforce. Build, secure, and scale without managing infrastructure","url":"https://www.cloudflare.com/","sameAs":["https://github.com/cloudflare","https://www.linkedin.com/company/cloudflare","https://x.com/cloudflare"],"logo":{"@type":"ImageObject","url":"https://developers.cloudflare.com/logo.svg"},"address":{"@type":"PostalAddress","streetAddress":"101 Townsend St","addressLocality":"San Francisco","addressRegion":"CA","postalCode":"94107","addressCountry":"US"},"contactPoint":[{"@type":"ContactPoint","contactType":"Customer Support","url":"https://support.cloudflare.com/","availableLanguage":["English"]},{"@type":"ContactPoint","contactType":"Sales","url":"https://www.cloudflare.com/contact/","availableLanguage":["English"]}]},"isPartOf":{"@type":"WebSite","@id":"https://developers.cloudflare.com/#website","name":"Cloudflare Docs","url":"https://developers.cloudflare.com/"}}

description: xAI's Grok 4.3 model with a 1M-token context window and strong agentic tool calling with minimal hallucinations. Accepts text and image inputs, and supports function calling, structured outputs, and configurable reasoning effort (none, low, medium, high). title: Grok 4.3 image: https://developers.cloudflare.com/og-docs.png

Documentation Index
Fetch the complete documentation index at: https://developers.cloudflare.com/llms.txt
Use this file to discover all available pages before exploring further.

Grok 4.20 Multi-Agent·续

Grok 4.20 Multi-Agent·续

Parameters

API Schemas (Raw)

On this page