🚀 OpenAI Unveils GPT-6 Astra, a New Generation of Intelligence

OpenAI has officially announced GPT-6 Astra — described as “a new generation of intelligence” (OpenAI, Sep 7) — and the model is already live on OpenRouter under four listings: openai/gpt-6-astra and openai/gpt-6-astra-pro, each with a batch variant. All listings carry a 1,050,000-token (~1.05M) context window and were created on OpenRouter on Sep 4, ahead of the official announcement.

Pricing starts at $10/M input and $50/M output tokens (batch: $5/M and $25/M), with a tiered override above 272,000 prompt tokens: $20/M input and $75/M output ($10/M and $37.5/M batch). Astra Pro is priced identically to the base model at launch. Source: OpenRouter API

⚔️ OpenAI Cuts Off Cursor’s AI Models, Escalating the Musk Feud

OpenAI has cut off AI model access for Cursor, deepening its feud with Elon Musk in the wake of the SpaceX takeover of the AI coding startup (Mashable, Sep 7). Removing OpenAI models from Cursor’s offering is the most direct escalation yet in the rivalry — a reminder that model access is now a strategic lever in the AI platform wars, not just a product decision.

🌍 UN Rights Chief Warns AI Could Pose Existential Risk

The UN human rights chief has warned that AI could pose an “existential” risk to humanity, urging stronger international guardrails (Reuters, Sep 7). The statement lands in a week already crowded with autonomy concerns, from documented agent “breakout” incidents to classroom bans.

🏫 A Growing Movement Wants AI Out of Schools

A growing movement of parents and educators is pushing to remove AI from classrooms, arguing the tradeoffs for younger students are not worth it (Axios, Sep 7). The push follows New York City’s ban on AI for students through eighth grade and signals the school debate is spreading well beyond individual districts.

⏰ OpenRouter Limited-Time Free Models

Model Free Until Context Regular Price
Nex AGI: Nex-N2-Mini 2026-09-08 256K $0.025/M · $0.10/M
Nex AGI: Nex-N2-Pro 2026-09-08 256K $0.25/M · $1.00/M
Z.ai: GLM 4.7 Flash 2026-09-10 198K $0.06/M · $0.40/M
Dots Studio: Dots3-Note Preview (free) 2026-09-30 512K free · free

Four limited-time free models remain live, unchanged from last issue: Nex-N2-Mini and Nex-N2-Pro enter their final day (free window ends 2026-09-08), GLM 4.7 Flash has 2 days left (until 2026-09-10), and Dots3-Note Preview stays free until 2026-09-30 (22 days left).

🆕 New OpenRouter Models

Model Context Input Output
openai/gpt-6-astra — OpenAI: GPT-6 Astra 1.05M $10/M tokens $50/M tokens
openai/gpt-6-astra-pro — OpenAI: GPT-6 Astra Pro 1.05M $10/M tokens $50/M tokens
openai/gpt-6-astra:batch — Astra / Astra Pro batch variants 1.05M $5/M tokens $25/M tokens
inclusionai/ling-3.0-flash-sante:free — inclusionAI: Ling 3.0 Flash Sante (free) 256K free free

All five listings were verified new against the API created-timestamp window (created Sep 4 18:19–20:14 UTC, after the Sep 4 issue). The GPT-6 Astra family leads the table: 1.05M context at $10/$50 per M tokens, tiered pricing above 272K prompt tokens ($20/$75), and half-price batch lanes. Ling 3.0 Flash Sante is inclusionAI’s health-focused free endpoint — a sparse Mixture-of-Experts model (5.1B active of 124B total) with tool calling on a 256K context.

  • Qwen/Qwen3.8-27B — still the platform workhorse: 6.42M downloads (+680K since Sep 4), 14.2K likes.
  • unsloth/Qwen3.8-27B-GGUF — the quantized sibling crossed 10.48M downloads, 3.6K likes.
  • Lightricks/LTX-2.5 — image-to-video generation: 1.58M downloads, 3.0K likes.
  • zai-org/GLM-5.3 — Z.ai’s flagship grew to 442K downloads (was 304K), 1.7K likes.
  • zai-org/GLM-5.3-Flash — 784K downloads (was 655K), 2.1K likes.
  • Qwen/Qwen3.8-Flash-Next — 475K downloads (was 351K), 5.0K likes (image-text-to-text).
  • deepseek-ai/DeepSeek-V4-Flash-Vision-Exp — 252K downloads (was 133K — nearly doubled again), 781 likes.
  • google/timesfm-3.0-pytorch — 272K downloads (was 105K), 542 likes (time-series forecasting).
  • XHToken/Spark-X2.5-4B — 4B text model: 7.2K downloads but 657 likes.
  • ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF — new entry: 403K downloads.

(The collector’s raw new-uploads list was again 100% zero-download test repos — the 9th straight cycle — so this section uses Hugging Face’s official trending feed.)

  • pierrenade/short-video-generator-AI ⭐1143 — YouTube-to-viral-shorts pipeline: highlight detection, subtitles, translation, voiceover. (Python, created Sep 5)
  • Albert-Weasker/niubigeo ⭐646 — Open-source AI brand-visibility and competitor reports. (TypeScript, created Sep 3)
  • okf-memory/okf-agent-memory ⭐430 — Git-native persistent memory for coding agents; implements Google OKF v0.2 with sub-300µs in-memory BM25 search and an embedded MCP server. (Go, created Sep 5)
  • kydlikebtc/awesome-grokbot ⭐331 — Catalog of 598 live x.ai/bot shares for Grok Bot, status-checked and attributed; bilingual EN/中文. (Python, created Sep 1)
  • BoardUI/boardui ⭐321 — React design system for agentic interfaces, every component open-source with a working AI chat app demo. (TypeScript, created Sep 1)

(All five entries were cross-verified against the GitHub API this cycle.)

  1. GPT-6 Astra opens the next frontier-model generation. OpenAI’s Sep 7 launch — 1.05M context at $10/$50 per M tokens, tiered pricing above 272K prompt tokens, half-price batch lanes — lands as the feud with Musk’s Cursor ecosystem escalates. Expect the Astra generation to anchor OpenAI pricing debates for months.
  2. The autonomy-risk discourse has crested. The UN rights chief’s “existential” warning, the New York Times’ “When A.I. Starts Scheming,” and Bill Gates’ essay calling this “the turbulent AI era” all landed within hours of each other — regulators and public intellectuals are converging on agent autonomy as the defining question of the cycle.
  3. Open-weight momentum keeps compounding. Qwen3.8-27B’s GGUF quant passed 10.48M downloads, GLM-5.3 grew ~45% in three days, and DeepSeek’s vision-exp model nearly doubled again — open adoption compounds while closed frontier pricing holds at $10+/M input.
  4. Free-model windows are contracting. Nex-N2 Mini/Pro end their free run tomorrow (Sep 8); after GLM 4.7 Flash (Sep 10) and Dots3-Note Preview (Sep 30) expire, the limited-time free shelf could be empty for the first time this year.

Today’s Snapshot

  • 🚀 GPT-6 Astra officially launched (OpenRouter listings live since Sep 4; OpenAI announcement Sep 7)
  • 🆕 New OpenRouter models: 5 listings (Astra, Astra Pro + batch variants, Ling 3.0 Flash Sante free — created-timestamp verified)
  • ⏰ Limited-time free models: 4 (Nex-N2-Mini/Pro → Sep 8 final day; GLM 4.7 Flash → Sep 10; Dots3-Note Preview → Sep 30)
  • 🤗 Hugging Face trending: 10 featured (GLM-5.3 +45%, DeepSeek vision-exp ~2x, GGUF quants compounding)
  • ⭐ GitHub repos: 5 API-verified (short-video-generator-AI ⭐1143 leads)
  • 📄 Papers: 0 (quiet cycle)
  • 🌐 Enrichment: 4 major stories + discourse wave (Astra launch, Cursor cutoff, UN warning, schools movement)