Context Collection Competition by JetBrains and Mistral AI

Build smarter code completions and compete for a share of USD 12,000!

In AI-enabled IDEs, code completion quality heavily depends on how well the IDE understands the surrounding code – the context. That context is everything, and we want your help to find the best way to collect it.

Join JetBrains and Mistral AI at the Context Collection Competition. Show us your best strategy for gathering code context, and compete for your share of USD 12,000 in prizes and a chance to present it at the workshop at ASE 2025.

Why context matters

Code completion predicts what a developer will write next based on the current code. Our experiments at JetBrains Research show that context plays an important role in the quality of code completion. This is a hot topic in software engineering research, and we believe it’s a great time to push the boundaries even further.

Goal and tracks

The goal of our competition is to create a context collection strategy that supplements the given completion points with useful information from across the whole repository. The strategy should maximize the chrF score averaged between three strong code models: Mellum by JetBrains, Codestral by Mistral AI, and Qwen2.5-Coder by Alibaba Cloud.

The competition includes two tracks with the same problem, but in different programming languages:

• Python: A popular target for many novel AI-based programming assistance techniques due to its very wide user base.
• Kotlin: A modern statically-typed language with historically good support in JetBrains products, but with less interest in the research community.

We’re especially excited about universal solutions that work across both dynamic (Python) and static (Kotlin) typing systems.

Prizes

Each track awards prizes to the top three teams: