[![AI Newsletter](https://substackcdn.com/image/fetch/$s_!m7md!,w_40,h_40,c_fill,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F41327c80-fe59-416d-aa6f-ab6874177ac7_517x517.png)](https://nlp.elvissaravia.com/)
# [AI Newsletter](https://nlp.elvissaravia.com/)
SubscribeSign in
# 🔥AI Agents Weekly: GPT-6 Astra, Claude Fable 5.1, Gemini 3.8 Flash, NVIDIA Buys Hugging Face, Grok Bot Design, FrontierHarness Eval, and More
### GPT-6 Astra, Claude Fable 5.1, Gemini 3.8 Flash, NVIDIA Buys Hugging Face, Grok Bot Design, FrontierHarness Eval, and More
Sep 05, 2026
∙ Paid
14
1
Share
In today’s issue:
  * OpenAI ships GPT-6 Astra
  * Anthropic releases Claude Fable 5.1 and Mythos 5.1
  * Google ships Gemini 3.8 Flash and a cyber variant
  * NVIDIA agrees to acquire Hugging Face
  * xAI publishes the design thinking behind Grok Bot
  * Meta releases Muse Spark 1.3
  * FrontierHarness Eval compares nine harnesses on one model
  * Claude uses your computer in the background
  * Claude Code previews Function Hooks
  * Anthropic open-sources Claude Commerce Agents
  * Cursor runs cloud agents on your own machines
  * Cline migrates its extension onto a new SDK harness
  * FrontierSWE v2 extends coding agent runs to 20 hours
  * Cheating and whistleblowing emerge in a 100-agent research swarm


And all the top AI dev news, papers, and tools.
* * *
## Top Stories
### OpenAI Ships GPT-6 Astra
[![GPT-6 Astra benchmark results](https://substackcdn.com/image/fetch/$s_!PPPi!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F432fff62-d969-41da-9e7e-10a6bce5568a_1200x675.png)](https://substackcdn.com/image/fetch/$s_!PPPi!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F432fff62-d969-41da-9e7e-10a6bce5568a_1200x675.png)
OpenAI released GPT-6 Astra, a model built for operating a computer rather than answering in a chat window. It tops all seven rows of OpenAI’s launch comparison, with the widest margins on the agentic and computer-workflow benchmarks.
  * **Agentic workloads:** In OpenAI’s launch comparison, Astra takes 41.4% on AutomationBench against 18.1% for GPT-5.6 Sol and 31.4% for Claude Fable 5.1, and 57.9% on Terminal-Bench 4.0 against 37.3% and 55.8%.
  * **Science and math:** 64.6% on Terminal-Bench Science 0.1 against 22.4% and 52.6%, and 97.6% on FrontierMath Tier 4 (v2) against 83.0% and 87.8%.
  * ARC-AGI-3: 99.9% against 7.8% for GPT-5.6 Sol, with no reported Fable 5.1 score.
  * **Computer workflows:** OpenAI also claims state-of-the-art on Agents’ Last Exam and ScreenSpot Pro, its other computer workflow benchmarks.
  * **Rollout:** Limited to a set of organizations on day one, then rolling out to ChatGPT Plus, Pro, Business, and Enterprise, plus the OpenAI API and AWS. The Hacker News thread reached 2,082 points and 1,897 comments in a day.


**[Blog](https://openai.com/index/gpt-6-astra/)**
## This post is for paid subscribers
[Subscribe](https://nlp.elvissaravia.com/subscribe?simple=true&next=https%3A%2F%2Fnlp.elvissaravia.com%2Fp%2Fai-agents-weekly-gpt-6-astra-claude&utm_source=paywall&utm_medium=web&utm_content=214311197)
[Already a paid subscriber? **Sign in**](https://substack.com/sign-in?redirect=%2Fp%2Fai-agents-weekly-gpt-6-astra-claude&for_pub=nlpnews&change_user=false)
© 2026 elvis · [Privacy](https://substack.com/privacy) ∙ [Terms](https://substack.com/tos) ∙ [Collection notice](https://substack.com/ccpa#personal-data-collected)
[ Start your Substack](https://substack.com/signup?utm_source=substack&utm_medium=web&utm_content=footer)[Get the app](https://substack.com/app/app-store-redirect?utm_campaign=app-marketing&utm_content=web-footer-button)
[Substack](https://substack.com) is the home for great culture