November 7–8 · San Francisco
Vals AI Hackathon
For one weekend, benchmaxxing is the point. Post-train a small model to the top of three Vals benchmarks, or build a new eval that frontier models still fail.
- 24
- Hours, noon to noon
- 150
- Spots
- 1–4
- People per team
- Free
- To attend
About
Train a model or build an eval.
Use the same infrastructure we use to benchmark model releases, including Valkyrie, our open-source harness for agentic evals. Top projects get featured on vals.ai.
Tracks
Post-Training
Post-train a small model and score it on three Vals benchmarks, using compute from our sponsors.
Ranked on a live leaderboard. Prizes for 1st, 2nd and 3rd.
Build Your Own Eval
Build a benchmark on Valkyrie for a task frontier models still get wrong. We cover model and compute costs.
Judged on demos. Prizes for 1st, 2nd and 3rd.
Schedule
Doors open in
Sat, Nov 7 · 12 PM PT
- Sat 12 PM Check-in and snacks
- Sat 1 PM Opening remarks
- Sat 2 PM Hacking begins
- Sat 5 PM Cybersecurity CTF (optional side quest)
- Sat 7 PM Dinner and speakers
- Sat 10 PM Dessert and show-and-tell
- Sun 9 AM Breakfast
- Sun 10 AM Submissions close, demos and judging
- Sun 12 PM Winners announced
FAQ
Who can apply?
Anyone can apply, but we give preference to applicants with technical experience.
Do I need a team?
No, but we recommend one. Apply with a team of up to four, or apply solo and we'll match you with a team on-site.
Are there prizes?
Yes. Prizes for 1st, 2nd and 3rd place in each track.
What do I get access to?
- Post-Training: Sponsor compute and credits, plus automatic scoring on all three benchmarks.
- Build Your Own Eval: Valkyrie, our open-source eval harness. We cover model and compute costs.
Is food provided?
Yes. Lunch, dinner, breakfast and snacks. Note any dietary restrictions when you apply.
Interested in sponsoring? Email rayan@vals.ai or miyuki@vals.ai.