←
Patrick Wendell1,136 posts
Patrick Wendell avatar✓

Patrick Wendell ✓

@pwendell

Co-founder and VP of engineering @databricks.

◷ Joined June 2009▧ 31 Photos and videos
Patrick Wendell avatar
Patrick Wendell@pwendell·Nov 16

A few people asked me to comment on why Astra/Fable usage seems quite low in this new dataset. I can't know for sure, but the #1 thing I hear from customers is that Opus/Sol/OSS models have simply gotten "good enough" for a huge fraction of enterprise work, especially when combined with routing. developers.databricks.com/ai…

Patrick Wendell avatar
Patrick Wendell@pwendell·Nov 16

5/ We've also published the internal benchmarks we use to measure the pareto frontier on cost-v-quality for techincal work. The Unity Gateway Smart Router leverages this information to chose task-appropriate models. The benchmark will be updated regularly.

Post attachment
Patrick Wendell avatar
Patrick Wendell@pwendell·Nov 16

4/ Spend controls: Most spend is incurred by power users (P99 spend is $700/week vs P50 is $2.50/week). Power user spend appears to be holding mostly flat, likely as a result of budgeting, while the median user spend is quickly growing from a small base.

Post attachment
Patrick Wendell avatar
Patrick Wendell@pwendell·Nov 16

3/ Rise of open source: OSS models are quickly increasing in token share amongst enterprises, growing from nearly zero to 30% in the 5 month period. Anecdotally, we see those models used mostly to reduce costs on high volume workloads (large batch workloads, coding, etc).

Post attachment
Patrick Wendell avatar
Patrick Wendell@pwendell·Nov 16

2/ Provider diversification: Enterprises are increasingly using models from more providers. At mid year most companies had one model provider. Today, 75% have two or more, and almost half use three or more. The most common "new providers" are OpenAI (GPT models) and OSS models.

Post attachment
Patrick Wendell avatar
Patrick Wendell@pwendell·Nov 16

1/ Today we’re publishing the Databricks AI Index, showing trends from enterprises who manage AI with our Unity Gateway. Data covers >10,000 companies with broad representation (Digital Natives, Fortune 500, Global 2000, etc...). Enterprise AI consumption is changing quite a bit:

Post attachment
Ali Ghodsi avatar
⟳ Patrick Wendell retweeted
Ali Ghodsi@alighodsi·Oct 28

I don't think we should fight. The two companies are complementary. Since you're not doing this AI stuff, I really think Snowflake should consider getting its AI tokens through Databricks. You can get both cheap OSS models as well as frontier (OAI/Ant/Gemini/Grok) models from our platform. It all goes through Unity Gateway which has Smart Routing, Budgets, and Access Control. You can even use Omnigent with any harness, including CoCo. Most of our customers used this to reduce token spend ~30% last few months while token count went up. Genie lets you then track, control, manage that spend, all your financials, marketing, etc. Quick POC?

Greg Brockman avatar
⟳ Patrick Wendell retweeted
Greg Brockman@gdb·Oct 26

Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory). AI starting to help us improve our understanding of the universe! openai.com/index/building-sc…

Jane Zhang avatar
⟳ Patrick Wendell retweeted
Jane Zhang@jjanezhang·Sep 30

🎉 Introducing Databricks AI Decide: make fast decisions on your governed data Following the TypeSafe AI Jev launch, we’ve seen increasing demand for a fast, low-cost API that turns raw text into structured decisions. However, many of our enterprise customers are unable to use Jev due to data privacy and access concerns. AI Decide is our enterprise-grade function for fast decisions on customer’s governed data. It is supported for batch use cases like processing millions of documents with SQL and for realtime applications like model routing through our REST API. Starting early next week, customers will also be able to govern permissions on the ai_decide function through Databricks Unity Gateway. A big shoutout to @mattydtweetz @ivanzhouyq @nihit_desai @hanlintang for getting this new function launched in less than a week!

Post attachment
Patrick Wendell avatar
Patrick Wendell@pwendell·Sep 29

We've moved to an approach @databricks that gives "Day 1" access to 12,000 employees whenever a new frontier model is released. Very tricky to do while avoiding massive cost spikes and quality regressions. We've written up our approach if others are interested (see below).

Post attachment
Ali Ghodsi avatar
⟳ Patrick Wendell retweeted
Ali Ghodsi@alighodsi·Sep 28

We just tested the latest AI models on 2,400 engineers. The cost/quality frontier just shifted massively. The TL;DR: • Opus 5.5: Our new daily default - higher quality and 20% cheaper than prior baselines. • GPT-6 Luna: Shockingly capable and 20x cheaper than Opus 5.5 for high-volume routing. Must-read for anyone managing production AI workloads 👇

Patrick Wendell avatar
Patrick Wendell@pwendell·Sep 28

Crazy few weeks for model releases! Our findings @Databricks show several new models meaningfully advance the pareto frontier. Results below (online workload analysis of N=2,400 engineers, plus offline evals): 1. Two of the three models released last week clearly expand the cost/quality frontier: Opus 5.5 and GPT-6 Luna. 2. Opus 5.5 is now the highest quality mid-tier model. It is better than all prior Opus models, better than GPT-6 Sol, and better than GPT-5.6 Sol. 3. Opus 5.5 reduces same-task costs consistently by 20% in both offline and online analysis. This is against a baseline of Opus 4.8, the prior least-cost Opus model (Opus 5.0 was a bit of a dud with high costs and barely noticeable quality improvements). 4. Due to best-in-class quality and lower costs, Opus 5.5 is a strong candidate as an “every day default” model for coding, and we are now encouraging it for this purpose at Databricks. 5. GPT-6 Luna is very, very, very cheap. It was at least 20 times cheaper per-task than Opus 5.5 in every offline benchmark we tested and in observed online use. 6. GPT-6 Luna is surprisingly capable given how cheap it is. On one of our most difficult evaluation suites it roughly matches Opus 4.6 performance, while being 99.3% cheaper per-task than Opus 4.6 was at that time. That's a 100X cost reduction in ~9 months! This finding is preliminary and we are still evaluating Luna quality on a broader set of offline and online tests. Our production setup: Unity Gateway to route workloads across models and trace agentic interactions. A mix of end-user harnesses including: Omingent (meta-harness), Claude Code, Codex, and Cursor.

Ali Ghodsi avatar
⟳ Patrick Wendell retweeted
Ali Ghodsi@alighodsi·Sep 24

All seven Databricks co-founders began our journey together at UC Berkeley. We started the company in a room in Soda Hall, moved to a small office on Addison Street, and spent our first few years in Berkeley before eventually heading to SF. Two of our co-founders are still on the faculty. Today we announced Databricks Field at California Memorial Stadium, our first collegiate athletics sponsorship. Berkeley shaped everything about how this company thinks, and this is our way of investing in the next generation of students and builders who'll do their best work on that campus. Go Bears! databricks.com/company/newsr…