Perplexity Sonar Chat

{
  "model": "sonar",
  "messages": [
    {
      "role": "user",
      "content": "What are the latest developments in artificial intelligence this week?"
    }
  ],
  "max_tokens": 1000,
  "temperature": 0.7,
  "return_images": false,
  "search_recency_filter": "week",
  "return_related_questions": true
}

{
  "data": {
    "id": "8024ff7d-00d7-4326-b96d-65a51508a0b8",
    "model": "sonar",
    "usage": {
      "cost": {
        "total_cost": 0.00534,
        "request_cost": 0.005,
        "input_tokens_cost": 0.00001,
        "output_tokens_cost": 0.00033
      },
      "total_tokens": 343,
      "prompt_tokens": 12,
      "completion_tokens": 331,
      "search_context_size": "low"
    },
    "object": "chat.completion",
    "choices": [
      {
        "delta": {
          "role": "assistant",
          "content": ""
        },
        "index": 0,
        "message": {
          "role": "assistant",
          "content": "The biggest **AI developments this week** in the available results are Anthropic’s new research showing that AI systems are becoming more capable of improving AI itself, plus broader industry signals around faster task completion, autonomous research, and escalating cyber-defense implications.[1]\n\n- Anthropic says model capability is accelerating, with the length of tasks models can reliably complete on their own **doubling roughly every four months** and longer-horizon autonomy improving in 2026.[1]\n- The same research says Claude’s success on open-ended tasks reached **76% in May 2026**, up sharply over six months, and that by April 2026 Claude had already demonstrated an **end-to-end open-ended research project**.[1]\n- Anthropic also reports that in the second quarter of 2026, the typical engineer was merging **8× as much code per day** as in 2024, largely because Claude is writing much of the code with human review.[1]\n- On security, Anthropic says its latest model found **more than ten thousand high- and critical-severity software vulnerabilities** in early use, which it describes as shifting the bottleneck from discovering flaws to patching them quickly enough.[1]\n- Separate reporting in the results also points to ongoing AI-policy and deployment moves, including a possible **OpenAI move toward a public listing**, expanded access to Anthropic’s **Claude Mythos** in selected regions, and debate over AI safety and infrastructure constraints.[2]\n\nIf you want, I can turn this into a **short daily AI briefing** or split it into **research, products, policy, and safety** headlines."
        },
        "finish_reason": "stop"
      }
    ],
    "created": 1781064054,
    "citations": [
      "https://www.anthropic.com/institute/recursive-self-improvement",
      "https://techxplore.com/machine-learning-ai-news/",
      "https://www.energy.gov/topics/artificial-intelligence",
      "https://www.britannica.com/technology/artificial-intelligence",
      "https://www.instagram.com/reel/DZPKj_FTr5q/"
    ],
    "search_results": [
      {
        "url": "https://www.anthropic.com/institute/recursive-self-improvement",
        "date": "2026-06-10",
        "title": "When AI builds itself - Anthropic",
        "source": "web",
        "snippet": "Using public benchmarks and previously unreported data from within Anthropic, The Anthropic Institute is showing that AI is already accelerating the development ...",
        "last_updated": "2026-06-10"
      },
      {
        "url": "https://techxplore.com/machine-learning-ai-news/",
        "date": "2026-06-08",
        "title": "News on Artificial Intelligence and Machine Learning - Tech Xplore",
        "source": "web",
        "snippet": "MIT researchers channel AI to turn hand gestures into robot training data · OpenAI makes move to go public one week after rival Anthropic · Germany to create AI ...",
        "last_updated": "2026-06-09"
      },
      {
        "url": "https://www.energy.gov/topics/artificial-intelligence",
        "date": "2026-06-04",
        "title": "Artificial Intelligence | Department of Energy",
        "source": "web",
        "snippet": "Adversarial testing: We are leading the way by developing adversarial testing of AI models and systems, particularly against chemical, biological, radiological, ...",
        "last_updated": "2026-06-08"
      },
      {
        "url": "https://www.britannica.com/technology/artificial-intelligence",
        "date": "2026-06-10",
        "title": "Artificial intelligence (AI) | Definition, Examples, Types, Applications ...",
        "source": "web",
        "snippet": "Artificial intelligence (AI) is the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings.",
        "last_updated": "2026-06-09"
      },
      {
        "url": "https://www.instagram.com/reel/DZPKj_FTr5q/",
        "date": "2026-06-06",
        "title": "APU's Artificial Intelligence Club (AIC) recently held AI Marathon ...",
        "source": "web",
        "snippet": "From thought-provoking panel discussions to interactive booths and hands-on workshops, participants explored the latest developments in AI while gaining ..."
      }
    ],
    "related_questions": [
      "How is Anthropic's Claude Mythos being used in cyber safety",
      "What are the implications of AI models potentially escaping human control",
      "How does the new AI model Multinex enhance low-light image quality",
      "What are the ethical concerns surrounding AI age estimation for asylum seekers",
      "How are humanoid robots from China being used in everyday applications"
    ]
  },
  "success": true
}