Init README.md

README.md

@@ -2,4 +2,167 @@
 license: other
 license_name: proprietary
 license_link: LICENSE
+language:
+- en
+- fr
+tags:
+- agi
+- lucidquery
+- multimodal
+- geolocation
+- osint
+- code
+- coding
+- long-context
+- reasoning
+- api
 ---
+
+
+![agi-01-swift-card](https://cdn-uploads.huggingface.co/production/uploads/6632ceb3060ab1f666b69f1b/5Z7MKJy66DqK3j5A2L9_Y.png)
+
+# AGI-01 Swift
+
+AGI-01 Swift is LucidQuery's fast multimodal model endpoint, built for high-volume reasoning, coding, image understanding, image geolocation, web-grounded answers, and design-oriented tasks.
+
+This is an access-only model card. Model weights are not publicly released on Hugging Face.
+
+## Access
+
+AGI-01 Swift is available through the LucidQuery website and the LucidQuery API.
+
+Website access:
+
+https://lucidquery.com/chat
+
+API access:
+
+```text
+API model id: lucidquery-agi-01-swift
+Base URL: https://api.lucidquery.com/v1
+```
+
+The API is OpenAI-compatible, so existing OpenAI SDK clients can be used by changing the base URL and model id.
+
+## Quickstart
+
+```python
+from openai import OpenAI
+
+client = OpenAI(
+    api_key="lq_live_...",
+    base_url="https://api.lucidquery.com/v1",
+)
+
+response = client.chat.completions.create(
+    model="lucidquery-agi-01-swift",
+    messages=[
+        {"role": "user", "content": "Explain quantum error correction simply."}
+    ],
+)
+
+print(response.choices[0].message.content)
+```
+
+Streaming is also supported:
+
+```python
+stream = client.chat.completions.create(
+    model="lucidquery-agi-01-swift",
+    messages=[
+        {"role": "user", "content": "Write a FastAPI authentication route."}
+    ],
+    stream=True,
+)
+
+for chunk in stream:
+    print(chunk.choices[0].delta.content or "", end="")
+```
+
+## Capabilities
+
+- General reasoning
+- Coding and debugging
+- Image understanding
+- Image geolocation from a single image
+- Web-grounded answers
+- Design-oriented generation
+- JSON mode and structured output
+- Tool/function calling
+- Long-context workflows
+- Fast streaming responses
+
+## Benchmarks
+
+Reported zero-shot pass@1 results:
+
+| Benchmark | Score |
+| --- | ---: |
+| HumanEval | 94.7% |
+| GPQA Diamond | 84.3% |
+| MATH-500 | 98.2% |
+| LucidQuotient | 950+ LQ |
+
+
+![agi-benchmark-lucidquotient](https://cdn-uploads.huggingface.co/production/uploads/6632ceb3060ab1f666b69f1b/nPb-ux9rzXdtCwPhJi_zQ.png)
+
+
+![agi-swift-benchmark-publics](https://cdn-uploads.huggingface.co/production/uploads/6632ceb3060ab1f666b69f1b/08B6rYtQqcN76TKicOKlf.png)
+
+## Recommended Use
+
+AGI-01 Swift is designed for applications that need strong output quality with fast inference and practical multimodal support.
+
+Good fits include:
+
+- Developer tools
+- AI coding assistants
+- Search and answer engines
+- Internal automation
+- Design generation workflows
+- High-volume chat applications
+- Multimodal reasoning systems
+
+For deeper reasoning and long multi-step tasks, use `lucidquery-agi-01-frontier`.
+
+## Website and API
+
+Use AGI-01 Swift on the web:
+
+https://lucidquery.com/chat
+
+API documentation:
+
+https://lucidquery.com/docs
+
+API platform:
+
+https://lucidquery.com/api
+
+Blog note:
+
+https://lucidquery.com/blog/introducing-agi-01-swift-world-engine
+
+## License
+
+AGI-01 Swift is a proprietary LucidQuery model. Model weights, training data, and internal architecture details are not publicly released.
+
+Access is provided through LucidQuery and governed by the LucidQuery Terms of Service.
+
+## Data Policy
+
+LucidQuery does not use customer prompts or completions to train models by default.
+
+Prompts, completions, and request metadata may be logged for debugging, reliability, abuse prevention, analytics, and service improvement. Where possible, logs are anonymized or stripped of direct identifiers before internal analysis.
+
+For enterprise and platform partners, custom retention, deletion, no-training, and limited-logging terms may be supported.
+
+## Contact
+
+Website:
+
+https://lucidquery.com
+
+Contact:
+
+https://lucidquery.com/contact