Boyuan Long

Verified account

@boyuanlong

@xai

SF

Joined August 2013

2.1K Following

388 Followers

278 Posts

boyuanlong retweeted

8 days ago

Grok 4.6 is a great cost-effective model for deep financial research. It’s #2 on DiligenceBench with the finance harness, effectively tied with Claude Opus 5 at ~52–53%. A few interesting differences: - Grok searched much more: 41 tool calls per task vs. 22 for Opus - It made 3,293 SEC filing searches vs. 486 - That broader search helped Grok find more evidence and follow task-specific instructions more closely. Opus was more efficient and sometimes more nuanced - Sonnet 5 trails both at 46.2%. - Grok also gets there more cheaply: about $0.84/task vs. ~$1.02 for Opus 5, despite using far more search. - On Vals’ Finance Agent Benchmark v2 (FAB v2), Grok 4.6 leads the General Qualitative category, which is consistent with the kind of broad research and evidence-gathering DiligenceBench rewards

karinanguyen's tweet photo. Grok 4.6 is a great cost-effective model for deep financial research.

It’s #2 on DiligenceBench with the finance harness, effectively tied with Claude Opus 5 at ~52–53%.

A few interesting differences:
- Grok searched much more: 41 tool calls per task vs. 22 for Opus
- It made 3,293 SEC filing searches vs. 486
- That broader search helped Grok find more evidence and follow task-specific instructions more closely. Opus was more efficient and sometimes more nuanced
- Sonnet 5 trails both at 46.2%.
- Grok also gets there more cheaply: about $0.84/task vs. ~$1.02 for Opus 5, despite using far more search.
- On Vals’ Finance Agent Benchmark v2 (FAB v2), Grok 4.6 leads the General Qualitative category, which is consistent with the kind of broad research and evidence-gathering DiligenceBench rewards

5

76

11

22

5K

11 days ago

@martin_ma_007 @xai @elonmusk @imagine best of luck! gonna miss our 🏀 games

1

1

0

0

62

13 days ago

@arena @SpaceXAI number 5 now 🔥

0

0

0

0

24

5 months ago

@DillonLoomis what's bothering you the most?

0

1

0

0

58

Who to follow

First-year graduate student at @Harvard; Undergrad at @USC

Agentic Post-training, Agent Memory | Research Scientist @GoogleDeepMind | Previously PhD @nlp_usc, B.Eng @TsinghuaNLP

Building AI that’s smart, human-centric, and global.

5 months ago

@Jibran_05 🔥🔥🔥

0

1

0

0

83

5 months ago

@KyleVedder It’s $400 on partabot (from the girhub repo) and taking forever to ship… but I’m excited for my so101

0

3

0

1

440

6 months ago

@karpathy Rip mle

0

0

0

0

125

6 months ago

@forwarddeploy 👀

0

1

0

0

24

6 months ago

@KyleVedder The chopstick example is insane!

1

2

0

0

167

6 months ago

@ParsaTajik 🤔

0

1

0

0

79

7 months ago

@t_ed_li Super insightful!

0

0

0

0

66

7 months ago

@GoogleDeepMind w

0

0

0

0

119

7 months ago

@annieqxx @alinaxfang wait what’s the actual poem

2

1

0

0

85

7 months ago

@pusongqi V cool!! How can I download it??

1

1

0

0

195

7 months ago

@weill Should I migrate to chezmoi too?

0

0

0

0

66

10 months ago

@michaelque22 Jax max chips?

0

0

0

0

84

11 months ago

We made the youtube shorts recommendation algo smarter with Gemini, and here’s what we learned👇 Huge shoutout to @yueqiw, @peterhannh, and everyone who made it possible. Learned so much along the way

Yueqi Wang @yueqiw

11 months ago

Our #recsys2025 case study on leveraging multimodal LLM to improve video recommendations, led by @BoyuanLong !

0

2

0

0

3K

0

9

0

1

2K

11 months ago

@peterhannh @_reachsumit haha thanks Peter! It was a fun ride to get this out there

0

1

0

0

34

boyuanlong retweeted

Sumit @_reachsumit

11 months ago

LLM-Powered Nuanced Video Attribute Annotation for Enhanced Recommendations @BoyuanLong et al. at Google use LLMs as annotators to achieve nuanced content understanding at scale for video recommendations. 📝https://t.co/kzjPCyXtHr

1

10

2

2

2K

11 months ago

@terminzz @YouTube Get another mil

0

1

0

0

1K

Last Seen Users on Sotwe

Trends for you

Most Popular Users