Phil Chen (@philhchen) / X

Phil Chen

375 posts

Phil Chen

@philhchen

Building something new. Previously research @openai @GoogleDeepMind @scale_AI @Stanford

Joined August 2022

Pinned
Phil Chen
@philhchen
Apr 21
I’ve started a new company with @tkkong! TK is a driving force behind a lot of Ramp’s success, building much of the core product, incubating the procurement platform, and leading Ramp Labs. We’re a team of IMO and Physics Olympiad gold medalists, and we’re hiring the most
TK Kong
@tkkong
Apr 21
I’ve started a new company with @philhchen! Phil built frontier LLMs across research & engineering at OpenAI, DeepMind, and Scale. I was shipping AI experiments at Ramp Labs. We've been heads down building personalized AI coworkers for every business. We’re growing our team
120K
Phil Chen
@philhchen
Feb 27, 2025
Excited to share that I’ve joined @OpenAI to work on GPT 4.57! Super excited to join an incredible team and continue to bring more intelligent models to the world
OpenAI
@OpenAI
Feb 27, 2025
GPT-4.5 has entered the Chat. openai.com/live/
68K
Phil Chen
@philhchen
May 18, 2024
Flash is natively multimodal, faster than Haiku/3.5, and supports 2M context. We’re adding caching coming soon for even faster latency.
Jeff Dean
@JeffDean
May 17, 2024
Gemini 1.5 Model Family: Technical Report updates now published In the report we present the latest models of the Gemini family – Gemini 1.5 Pro and Gemini 1.5 Flash, two highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information
89K
Phil Chen
@philhchen
Feb 4, 2025
One of the best resources for learning about modern ML systems for LLMs
Jacob Austin
@jacobaustin132
Feb 4, 2025
Making LLMs run efficiently can feel scary, but scaling isn’t magic, it’s math! We wanted to demystify the “systems view” of LLMs and wrote a little textbook called “How To Scale Your Model” which we’re releasing today. 1/n
5.3K
Phil Chen
@philhchen
Dec 11, 2024
Native multimodal generation is AGI
Tim Brooks
@_tim_brooks
Dec 11, 2024
Gemini 2.0 Flash has native image outputs! Congrats to the awesome team that built it. I find the example at 1:15 super cool: to change the car's color and add beach gear, the model generates two images step-by-step using visual chain of thought. youtube.com/watch?v=7RqFLp…
4K
Phil Chen
@philhchen
May 21, 2024
Latency, cost, and quality: choose all three with flash
Ben (no treats)
@andersonbcdefg
May 20, 2024
gemini-1.5 flash was the winner (by a substantial margin!) - claude haiku is a yes-man :( he told me a shitty classifier was nearly 100% accurate - 3.5-turbo errs on the side of false negatives: impossible to please! - flash is the most balanced & has highest agreement with gpt4o
2.5K
Phil Chen
@philhchen
May 18, 2024
Having worked both as a heavy GCP user and as an engineer on GCP, I’ve been pretty frustrated with both reading bad docs and spending a lot of time thinking about how to write accurate but succinct docs. Extremely bullish on LLMs to solve this, especially with cheap long context
4.4K
Phil Chen
@philhchen
Nov 20, 2024
Personal highlights from working with Ema and the video: - Incredibly deep understanding of the full stack, from network and chip design to retrieval and ML algos - Bringing people together from all across the company to run, make espressos, and ship together - Volunteering to
Logan Kilpatrick
@OfficialLoganK
Nov 19, 2024
Gemini 1.5 Flash-8B is becoming one of the most widely used LLMs in the world! In my conversation with Google Fellow Emanuel Taropa, we talk all things Flash-8B & what it takes to move quickly at Google 🚢 Check out the full episode below and on YouTube: youtube.com/watch?v=SIVoyL…
00:00
1.4K
Phil Chen
@philhchen
Sep 16, 2024
Fastest and cheapest intelligence at scale
Logan Kilpatrick
@OfficialLoganK
Sep 16, 2024
We just shipped a series of changes which have significantly improved the Gemini 1.5 Flash latency (>3x reduction) and output tokens per second (>2x more)⚡️🚢
2K
Phil Chen
@philhchen
Dec 6, 2024
Intelligence too cheap to meter
Jeff Dean
@JeffDean
Dec 6, 2024
Today’s the one year anniversary of our first Gemini model releases! And it’s never looked better. Check out our newest release, Gemini-exp-1206, in Google AI Studio and the Gemini API! aistudio.google.com/app/prompts/ne…
1.3K
Phil Chen
@philhchen
Dec 19, 2024
A preview of what we've been working on - incredibly fun, thoughtful model to play with, along with fully transparent thoughts for introspection
Noam Shazeer
@NoamShazeer
Dec 19, 2024
We’ve been *thinking* about how to improve model reasoning and explainability Introducing Gemini 2.0 Flash Thinking, an experimental model trained to think out loud, leading to stronger reasoning performance. Excited to get this first model into the hands of developers to try
1.7K
Phil Chen
@philhchen
Dec 28, 2024
Replying to @martin_casado
RL with formal verification + critic model could cover most of formal education, which relies partially on assigning grades (aka rewards) to assignments and exams. Even if grades on essays aren’t formally verified for consistency, they’re still noisy signal that humans learn from
939
Phil Chen
@philhchen
Sep 20, 2024
Incredible work from the team on long context evals beyond needle in the haystack
Kiran Vodrahalli
@kiranvodrahalli
Sep 20, 2024
Happy to share Michelangelo (arxiv.org/abs/2409.12640), a long-context reasoning benchmark which measures performance beyond needle tasks up to arbitrary context lengths and remains challenging for frontier models. Stay tuned for more Michelangelo evals to come!
862
Phil Chen
@philhchen
Nov 30, 2024
Replying to @karpathy
Inference time compute models like o1 aren’t really a direct imitation of human data labelers. RL over long sequences, rewarded by automatic non-human feedback (e.g. code execution or math answer checking), can pretty dramatically change the output distribution.
359