**Vals.ai Benchmark – International Olympiad in Informatics**

> Informatics olympiad programming problems Key Takeaways Unlike the saturated knowledge benchmarks, IOI still sharply separates models: GPT-6 Astra solves every problem in all three years, GPT-5.6 Sol (91.17%), Claude Fable 5.1 (90.78%) and GPT-5.6 Terra (87.61%) follow, and the twenty-six-model field then spreads across ninety points, down to 9.33%, so competitive-programming ability remains a real differentiator. The agent sees only what a contestant sees: the statement, the sample grader and one sample. It has no tests, no internet and no submission feedback, so every point comes from code the model wrote and checked itself. Scores decline on the newest problems: cohort mean accuracy is 58.38% on 2024 and 53.67% on 2025 but 50.82% on 2026, and fifteen of the twenty-six models score lowest on the 2026 problems. GPT 5.3 Codex drops from 62.83% on 2024 to 41.67% on 2026. The two leaders score 100% on 2026 even though the contest ran after their training cutoffs, so recency alone does not cap performance. Why IOI? Recently, top LLM labs like OpenAI and Google reported that their models achieved gold medals on the International Mathematical Olympiad (IMO). However, advanced models are starting to saturate IMO, meaning it may no longer effectively differentiate between the capabilities of top-performing models. Reports also suggest the evaluation process faced coordination challenges, with AI companies seeking expedited validation mid-competition that may not reflect standard IMO assessment procedures. The International Olympiad in Informatics (IOI) offers several advantages as an LLM benchmark. Unlike the IMO, the IOI is not yet saturated, providing clear differentiation between model capabilities. The competition features standardized and automated grading, ensuring objective evaluation without subjective scoring. Additionally, the IOI has real-world relevance as it tests C++ programming skills that are directly applicable to software development. Benchmark Design We designed our benchmark to imitate competition conditions as closely as possible. Agent Harness Each model runs inside OpenCode , a general-purpose coding agent, in an isolated sandbox with a c++ (v20) toolchain. Its workspace holds exactly what the contest hands a contestant : the problem statement, the task header and solution stub, the sample grader or manager, the compile and run scripts, and the sample input and output from the statement. The official test data, the subtask test lists, and the grading script are withheld until the attempt is over. The agent reads, writes, compiles and tests files freely in its workspace, and writes its answer to /workspace/solution.cpp . Like a contestant, it has no access to the public internet: the sandbox reaches only our model gateway, so published editorials and reference solutions are out of reach. Unlike a contestant, the agent has no submission tool and therefore receives no score feedback during the attempt: it can run the samples and whatever tests it writes for itself, and must decide on its own when its solution is good enough. Some models spend their entire per-response output budget reasoning about a problem before writing any code. When that happens, the harness continues the cut-off turn on the next step rather than failing the problem: it replays the provider’s own reasoning state where the API returns one, and otherwise restarts the turn with an instruction to work in smaller steps. The same rule applies to every model on this board, and the cost column includes the reasoning spent this way. Results from our earlier harness, which gave the agent an interactive grading tool, are preserved on IOI v1 . The two harnesses are not comparable, so their results are reported on separate boards. Scoring Grading matche...

**来源信息**
- **来源**：Hacker News：AI 热帖
- **分类**：论文
- **发布时间**：2026-09-12 23:05（北京时间）
- **原文**：[打开原文](https://www.vals.ai/benchmarks/ioi)