DeepReinforce's GrandCode Beats Every Human in Three Live Coding Contests DeepReinforce's multi-agent reinforcement learning system GrandCode placed first in three consecutive live Codeforces contests — Round 1087 on March 21, Round 1088 on March 28, and Round 1089 on March 29, 2026 — beating every human competitor including grandmasters, according to the company's technical report on arXiv. GrandCode coordinates a hypothesis proposer, main solver, test generator, and summarizer on a Qwen 3.5 mixture-of-experts backbone, trained with a new Agentic GRPO algorithm that handles delayed rewards and off-policy drift in long agent rollouts. The results compare with OpenAI o3's 175th place in April 2025 and Gemini 3.1 Pro's 8th place in February 2026, though DeepReinforce notes differing model families, compute budgets, and evaluation conditions prevent a controlled comparison; the code is released on GitHub. DeepReinforce's GrandCode Beats Every Human in Three Live Coding Contests A multi-agent reinforcement learning system from DeepReinforce swept three consecutive live Codeforces rounds, outperforming every human including legendary grandmasters. - GrandCode won Codeforces Rounds 1087, 1088, and 1089 https://arxiv.org/abs/2604.02721?utm source=alphasignal , beating every human including grandmasters. - Multi-agent system: hypothesis proposer, solver, summarizer, and adversarial test generator, built on Qwen 3.5 MoE. - New Agentic GRPO algorithm handles delayed rewards and off-policy drift in long agent rollouts. - Prior best AI results: Gemini 3.1 Pro placed 8th, OpenAI o3 placed 175th a year earlier. - Training uses seed problems from TACO, USACO, CodeContests, IOI, expanded with Gemini and Claude-generated traces. - Code released on GitHub https://github.com/deepreinforce-ai/codeforces?utm source=alphasignal ; approach generalizes best to verifier-rich coding tasks. DeepReinforce reports that GrandCode, its multi-agent reinforcement learning system, placed first in three consecutive live Codeforces contests: Round 1087 on March 21, Round 1088 on March 28, and Round 1089 on March 29, 2026. The results and system design appear in the project’s technical report https://arxiv.org/abs/2604.02721?utm source=alphasignal , and the code is available in a GitHub repository https://github.com/deepreinforce-ai/codeforces?utm source=alphasignal . Competitive programming has remained difficult for coding models because success requires algorithm design, exact implementation, and rapid verification against hidden tests. GrandCode’s reported wins suggest that specialized agents, automated testing, and online policy updates can outperform strong human contestants within that tightly defined environment. A steep rise under different conditions The paper compares GrandCode with earlier Codeforces results from OpenAI and Google systems. Those placements indicate rapid progress, although differences in model families, compute budgets, and evaluation conditions prevent a controlled comparison. | Codeforces placements cited by DeepReinforce | | | | |---|---|---|---| | System | Date | Placement | Evaluation context | |---|---|---|---| | OpenAI o3 | April 2025 | 175th | Earlier Codeforces evaluation | | Gemini 3.1 Pro | February 2026 | 8th | Non-live evaluation | | GrandCode | March 2026 | 1st in three rounds | Live contests, according to the paper | Why Codeforces resists coding models Codeforces problems combine original algorithm design with strict time, memory, and correctness constraints. Partial credit is uncommon, and a single unhandled edge case can invalidate an otherwise sound submission. Contestants typically have two or three hours to derive an approach, prove its complexity, implement it, and survive adversarial hidden tests. Codeforces ratings also reflect deep specialization. Ratings from 2400 to 2599 correspond to Grandmaster, 2600 to 2999 to International Grandmaster, and 3000 or higher to Legendary Grandmaster. Finishing first means outperforming contestants who have spent years practicing this specific form of problem solving. Conventional coding-model pipelines often fail before implementation because many contest problems depend on one structural or mathematical insight. Exploring every plausible approach creates an enormous search tree, and a candidate may appear correct until an unseen input exposes a flaw. Four specialists close the loop GrandCode coordinates a hypothesis proposer, main solver, test generator, and summarizer. Its main solver uses a Qwen 3.5 mixture-of-experts backbone, which activates a subset of model parameters for each token. DeepReinforce improves the components through post-training and online test-time reinforcement learning. This story is for Pro members You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.