
What happened
OpenAI's GPT-6 Astra is now in private preview on Snowflake Cortex AI, as a launch partner. It sets new highs in computer use, software engineering, and scientific reasoning over GPT-5.6 Sol.
Why it matters
GPT-6 Astra outperforms GPT-5.6 Sol on every listed benchmark, e.g., 97.6% vs 83.0% on FrontierMath Tier 4. It also uses about 65% fewer output tokens than Claude Opus 5 on Agents' Last Exam.
What to watch
Astra's alignment is notable: 0% unauthorized actions in a no-safeguards test, vs 48% for GPT-5.6 Sol. Watch for broader enterprise adoption as private preview expands.
WHO IT HITSEnterprise teams using Snowflake for coding, data engineering, and knowledge work can now leverage more powerful and aligned AI models. Data engineers and analysts may build production pipelines faster, while knowledge workers gain advanced agentic helpers.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
GPT-6 Astra represents a major step forward for OpenAI, combining advances in reasoning, computer use, and alignment. The launch is tightly integrated with Snowflake's enterprise platform, making frontier AI capabilities accessible within a governed environment. For Snowflake customers, this means they can apply these models to complex tasks—like long-running agentic workflows or software engineering—while maintaining security and compliance.
The performance benchmarks highlight Astra's superiority over GPT-5.6 Sol across a wide range of tasks, from research-level math to desktop computer use. Notably, Astra uses fewer resources, as seen in its reduced output token count and faster task completion, which could lead to lower costs and higher efficiency for enterprises. The alignment improvements, showing zero unauthorized actions in a stress test, are particularly significant for organizations with strict governance requirements.
As private preview rolls out, the true test will be how well these capabilities translate into real-world productivity gains within Snowflake's ecosystem. Success will depend on whether teams can effectively integrate Astra into their workflows with CoCo and CoWork, and whether the promised security and governance features hold up. Snowflake's role as a launch partner suggests this is only the beginning of a deeper collaboration.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Michael Burry, famous for The Big Short, is short Nvidia, Palantir and Tesla, and in his Substack newsletter s…

Investors have three creative routes to Anthropic exposure before its expected IPO: buying Alphabet, Amazon, o…

DeepSeek launched V4.1-Flash, a 763B-parameter open-weight model with a causal encoder-decoder architecture

Much of the attention on AI infrastructure buildouts is now tied to sheer compute power, with dominance define…

Barron's reported September 10 that Kepler Computing emerged from stealth with a memory architecture using fer…

Dynatrace acquired Arize AI, adding AI observability, evaluation and agent monitoring to its application obser…