←
Boris Power4,028 posts
Boris Power avatar

Boris Power

@BorisMPower

Head of Applied Research @OpenAI

◷ Joined July 2017▧ 234 Photos and videos
Boris Power avatar
Boris Power@BorisMPower·22 Aug 2025

At @OpenAI, we believe that AI can accelerate science and drug discovery. An exciting example is our work with @RetroBiosciences, where a custom model designed improved variants of the Nobel-prize winning Yamanaka proteins. Today we published a closer look at the breakthrough. ⬇️

Post attachment
Haider. avatar
⟳ Boris Power retweeted
Haider.@haider1·19h

openai models are the least benchmaxed among all labs the efficiency gain with astra and sol 6.2 is massive i had said before that this was oai’s real edge right now and they seem to be doubling down on it -- about 6x cheaper per task than fable 5.2 mostly because it uses nearly 7x fewer tokens

Post attachment
Jai avatar
⟳ Boris Power retweeted
Jai@Laneless_·21h

Which way, Western models?

Post attachment

We also evaluated how many serial reasoning steps GPT-6.2 Astra can do in a single output token (without chain of thought). It can perform 12.3 steps, versus 10.5 for GPT-6 Astra and 8.6 for GPT-6.1 Sol, the previous highs. We measure the number of steps models can perform at an accuracy of 25% above chance, where each step is normalized for complexity. Before GPT-6 Astra’s release, the most capable model was Claude Fable 5.1, which is able to perform only 4.2 serial steps. The oldest model we assessed is Claude Opus 3, which performed 2.3 steps. This capability is notable because models that can do complex reasoning without any chain of thought may be harder to monitor. We hope to finalize this and other measurements of more abstract model capabilities, and release them in the coming weeks.

Post attachment
Sherwin Wu avatar
⟳ Boris Power retweeted
Sherwin Wu@sherwinwu·Nov 17

The new Astra for Law (built on 6.2 Astra) leads on legal intelligence! Now available for all ChatGPT Enterprise customers via the Trusted Access Program. openai.com/index/astra-law-6… Absolutely GOATed work by the research team to get the pareto curve to look like this:

Post attachment
Boris Power avatar
Boris Power@BorisMPower·18h

Great models and super cheap too, dominating the pareto!

GPT-6.2 Astra from @OpenAI on ARC-AGI (Verified): - ARC-AGI-3: 100.0% (SOTA), $4.1K (provider adapter), 70.3%, $24.2K (standard) - ARC-AGI-2: 95.8% (SOTA), $0.50/task - ARC-AGI-1: 98.5%, $0.17/task GPT-6.2 Astra is the first model to achieve a 100% score on ARC-AGI-3.

Post attachment
Sherwin Wu avatar
⟳ Boris Power retweeted
Sherwin Wu@sherwinwu·Nov 17

Really proud that the new Astra for Law (and 6.2 Astra itself) leads the industry in eliminating hallucinations: 20x fewer hallucinations than Opus 5.5, and 25x fewer than Fable 5.2! Especially important for law firms, which are particularly sensitive about hallucinations.

Post attachment
OpenAI avatar
⟳ Boris Power retweeted
OpenAI@OpenAI·Nov 17

GPT-6.2 Astra—a frontier model built for your hardest projects. It sets new highs on evals across professional work and scientific research at less than a third of the cost per task of GPT‑6 Astra. You’ll see the difference in things like slideshows, websites, and 3D designs.

Post attachmentPost attachmentPost attachment
Boris Power avatar
Boris Power@BorisMPower·Nov 17

GPT-6.2 Astra scores 5.5 points higher on agents’ last exam, while 20x cheaper than the best Anthropic model! Wild 🤯

OpenAI@OpenAINov 17

Some of you have been asking very politely. Some of you have just been asking. GPT-6.2 Astra and Sol are now available. Super efficient and built to perform where it counts—the real world. openai.com/index/introducing…

1:59
Boris Power avatar
Boris Power@BorisMPower·Nov 12

Very interesting new era for archeological discovery!

tee@Tee201_Nov 12

In light of the viral post yesterday by @PietroMiotti, I am sharing 20+ verified archaeological discoveries made by ai agents. I started this project over 2 months ago out of passion for doing meaningful, important work with ai. Despite the community note confirming that Pietro's discovery was actually known and recorded, his post opened a door for archaeology. github.com/tee-20190/lidar-d… provides everything you need to make discoveries yourself, including proof of these discoveries by archaeologists. Today we're publishing the full record: every outcome, the negatives, and every correction. 🧵

Post attachment
Jeffrey Wang avatar
⟳ Boris Power retweeted
Jeffrey Wang@jeffreygwang·Nov 11

(1/10) Today, we’re sharing new research from @OpenAI: How does CoT monitorability (no-CoT capability and CoT controllability) evolve over the model training lifecycle? We emerged with unsurprising but sobering findings: as we scale toward more intelligent systems, models are becoming harder to monitor through their CoT. 🧵

Post attachment
Boris Power avatar
Boris Power@BorisMPower·Nov 11

This necessitates a more automated way to address these quickly

"When the company ran diagnostics on about 700 systems, it found 10,500 vulnerabilities and held an emergency press conference on June 16, applying fixes over three months. However, when it re-ran diagnostics using "GPT-6 Astra," the latest model released by US-based OpenAI, it confirmed 35,000 new vulnerabilities. "Honestly, we're completely exhausted" finance.biggo.com/news/2cb8f…

Boris Power avatar
Boris Power@BorisMPower·Nov 11

“Hotel” and “girl” are dangerously close when swipe typing on an iPhone. A word of caution ;)

Boris Power avatar
Boris Power@BorisMPower·Nov 11

A glimpse of an exciting future where code is provably correct for all critical services!

Delighted to share the the work of my team at @GoogleDeepMind: formally verifying a real-world implementation of the #Signal cryptographic protocol at the machine-code level. The proof was done using AlphaProof Code, our new AI framework for formal specification and verification.