# AI News Digest - 2026-07-20

1. Alibaba unveiled Qwen 3.8, a multimodal AI model with 2.4 trillion parameters, made an open-weight preview available, and stated the model rivaled leading systems while trailing only Fable 5.

2. Moonshot's Kimi K3 topped the Code Arena: Frontend rankings, outperforming Claude Fable 5 and GPT-5.6 Sol, but scored about 39% on FrontierMath Tier 4 compared with roughly 90% for models from OpenAI and Anthropic.

3. Google DeepMind repurposed a video generator in GenCeption to perform classic computer vision tasks such as depth estimation and segmentation, matching state-of-the-art systems while training on far less data and using predominantly synthetic videos.

4. RadLE 2.0 benchmark showed many AI models for radiology produced incorrect findings with full confidence, and human radiologists remained substantially more accurate.

5. Epoch AI tested three leading AI text detectors—Pangram, GPTZero, and Originality.ai—and found up to 18% of AI-generated passages went undetected overall and up to 48% for scientific writing.

# References

1. [https://the-decoder.com/alibabas-qwen-takes-on-kimi-k3-with-open-weight-qwen-3-8-says-model-is-second-only-to-fable-5/](https://the-decoder.com/alibabas-qwen-takes-on-kimi-k3-with-open-weight-qwen-3-8-says-model-is-second-only-to-fable-5/)

[Alibaba's Qwen takes on Kimi K3 with open-weight Qwen 3.8, says model is "second only to Fable 5"](https://the-decoder.com/alibabas-qwen-takes-on-kimi-k3-with-open-weight-qwen-3-8-says-model-is-second-only-to-fable-5/)

1. [https://the-decoder.com/moonshots-kimi-k3-outperforms-fable-5-in-frontend-code-but-lags-far-behind-in-complex-math/](https://the-decoder.com/moonshots-kimi-k3-outperforms-fable-5-in-frontend-code-but-lags-far-behind-in-complex-math/)

[Moonshot's Kimi K3 outperforms Fable 5 in frontend code but lags far behind in complex math](https://the-decoder.com/moonshots-kimi-k3-outperforms-fable-5-in-frontend-code-but-lags-far-behind-in-complex-math/)

1. [https://the-decoder.com/google-deepmind-argues-video-generators-already-contain-the-world-models-computer-vision-has-been-missing/](https://the-decoder.com/google-deepmind-argues-video-generators-already-contain-the-world-models-computer-vision-has-been-missing/)

[Google Deepmind argues video generators already contain the world models computer vision has been missing](https://the-decoder.com/google-deepmind-argues-video-generators-already-contain-the-world-models-computer-vision-has-been-missing/)

1. [https://the-decoder.com/ai-chatbots-reading-x-rays-can-be-dangerously-confident-even-when-theyre-wrong/](https://the-decoder.com/ai-chatbots-reading-x-rays-can-be-dangerously-confident-even-when-theyre-wrong/)

[AI chatbots reading X-rays can be dangerously confident even when they're wrong](https://the-decoder.com/ai-chatbots-reading-x-rays-can-be-dangerously-confident-even-when-theyre-wrong/)

1. [https://the-decoder.com/ai-text-detectors-struggle-when-language-models-mimic-an-authors-style/](https://the-decoder.com/ai-text-detectors-struggle-when-language-models-mimic-an-authors-style/)

[AI text detectors struggle when language models mimic an author's style](https://the-decoder.com/ai-text-detectors-struggle-when-language-models-mimic-an-authors-style/)

For the site tree, see the [root Markdown](https://ixtj.dev/.md).