{"benchmark": {"slug": "hal-usaco", "name": "USACO (HAL)", "description": "Competitive programming benchmark (USACO) on HAL.", "category": "eval", "split": null, "language": null, "task_count": 135, "homepage_url": "https://hal.cs.princeton.edu/usaco", "paper_url": "https://arxiv.org/abs/2510.11977", "hf_dataset_id": null, "official_leaderboard_url": "https://hal.cs.princeton.edu/usaco", "is_public_leaderboard": true, "harness_count": 2, "model_count": 11, "score_count": 12, "best_success_rate": 69.06}, "protocol": {"eval_protocol": "HAL USACO (HAL) accuracy", "temperature": null, "max_turns": null, "context_budget_tokens": null, "n": 12}, "rankings": {"benchmark": {"slug": "hal-usaco", "name": "USACO (HAL)", "description": "Competitive programming benchmark (USACO) on HAL.", "category": "eval", "split": null, "language": null, "task_count": 135, "homepage_url": "https://hal.cs.princeton.edu/usaco", "paper_url": "https://arxiv.org/abs/2510.11977", "hf_dataset_id": null, "official_leaderboard_url": "https://hal.cs.princeton.edu/usaco", "is_public_leaderboard": true, "harness_count": null, "model_count": null, "score_count": null, "best_success_rate": null}, "sort": "success_rate", "rows": [{"rank": 1, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gpt-5-medium", "name": "GPT-5 (medium)", "vendor": "OpenAI"}, "success_rate": 69.06, "resolved_count": null, "total_count": null, "cost_usd": 0.8639, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 2, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "o4-mini", "name": "o4 mini", "vendor": "openai"}, "success_rate": 53.09, "resolved_count": null, "total_count": null, "cost_usd": 0.1822, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 3, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "success_rate": 51.47, "resolved_count": null, "total_count": null, "cost_usd": 1.9831, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 4, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-4-1-opus", "name": "Claude 4.1 Opus", "vendor": "Anthropic"}, "success_rate": 48.21, "resolved_count": null, "total_count": null, "cost_usd": 2.0459, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 5, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "o3", "name": "o3", "vendor": "openai"}, "success_rate": 46.25, "resolved_count": null, "total_count": null, "cost_usd": 0.4244, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 6, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gpt-4-1", "name": "GPT-4.1", "vendor": "openai"}, "success_rate": 44.95, "resolved_count": null, "total_count": null, "cost_usd": 0.2081, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 7, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "deepseek-v3-march-2025", "name": "DeepSeek V3 (March 2025)", "vendor": "DeepSeek"}, "success_rate": 39.09, "resolved_count": null, "total_count": null, "cost_usd": 0.0895, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 8, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "deepseek-r1-january-2025", "name": "DeepSeek R1 (January 2025)", "vendor": "DeepSeek"}, "success_rate": 38.11, "resolved_count": null, "total_count": null, "cost_usd": 0.5929, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 9, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-3-7-sonnet", "name": "Claude 3.7 Sonnet", "vendor": "Anthropic"}, "success_rate": 29.32, "resolved_count": null, "total_count": null, "cost_usd": 0.2867, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 10, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gemini-2-0-flash", "name": "Gemini 2.0 Flash", "vendor": "Google"}, "success_rate": 27.04, "resolved_count": null, "total_count": null, "cost_usd": 0.0108, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 11, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-3-7-sonnet-high", "name": "Claude 3.7 Sonnet (high)", "vendor": "Anthropic"}, "success_rate": 26.71, "resolved_count": null, "total_count": null, "cost_usd": 0.418, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}, {"rank": 12, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gpt-4-1", "name": "GPT-4.1", "vendor": "openai"}, "success_rate": 25.41, "resolved_count": null, "total_count": null, "cost_usd": 1.4617, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null}]}}