{"benchmark": {"slug": "hal-swe-bench-verified-mini", "name": "SWE-bench Verified Mini (HAL)", "description": "HAL's SWE-bench Verified Mini board evaluates community scaffolds on a 49-task subset of Verified issue-fixing tasks. Catalog ingests HTML leaderboard rows (`hal_site`).\n\nCompare against swebench.com Verified (500 tasks) and bash-only mini-SWE-agent rows only when explaining subset and scaffold effects \u2014 not as interchangeable rankings.", "category": "eval", "split": null, "language": "python", "task_count": 49, "homepage_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "paper_url": "https://arxiv.org/abs/2510.11977", "hf_dataset_id": null, "official_leaderboard_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "is_public_leaderboard": true, "harness_count": 2, "model_count": 17, "score_count": 31, "best_success_rate": 72.0}, "protocol": {"eval_protocol": "HAL SWE-bench Verified Mini (HAL) accuracy", "temperature": null, "max_turns": null, "context_budget_tokens": null, "n": 31}, "rankings": {"benchmark": {"slug": "hal-swe-bench-verified-mini", "name": "SWE-bench Verified Mini (HAL)", "description": "HAL's SWE-bench Verified Mini board evaluates community scaffolds on a 49-task subset of Verified issue-fixing tasks. Catalog ingests HTML leaderboard rows (`hal_site`).\n\nCompare against swebench.com Verified (500 tasks) and bash-only mini-SWE-agent rows only when explaining subset and scaffold effects \u2014 not as interchangeable rankings.", "category": "eval", "split": null, "language": "python", "task_count": 49, "homepage_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "paper_url": "https://arxiv.org/abs/2510.11977", "hf_dataset_id": null, "official_leaderboard_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "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": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "claude-4-5-sonnet-high", "name": "Claude 4.5 Sonnet (high)", "vendor": "Anthropic"}, "success_rate": 72.0, "resolved_count": null, "total_count": null, "cost_usd": 9.4673, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 2, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "claude-4-5-sonnet", "name": "Claude 4.5 Sonnet", "vendor": "Anthropic"}, "success_rate": 68.0, "resolved_count": null, "total_count": null, "cost_usd": 10.3249, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 3, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "claude-4-1-opus", "name": "Claude 4.1 Opus", "vendor": "Anthropic"}, "success_rate": 61.0, "resolved_count": null, "total_count": null, "cost_usd": null, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 4, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "claude-3-7-sonnet-high", "name": "Claude 3.7 Sonnet (high)", "vendor": "Anthropic"}, "success_rate": 54.0, "resolved_count": null, "total_count": null, "cost_usd": 7.9363, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 5, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "success_rate": 54.0, "resolved_count": null, "total_count": null, "cost_usd": null, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 6, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "o4-mini", "name": "o4 mini", "vendor": "openai"}, "success_rate": 50.0, "resolved_count": null, "total_count": null, "cost_usd": 5.0706, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 7, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "claude-3-7-sonnet", "name": "Claude 3.7 Sonnet", "vendor": "Anthropic"}, "success_rate": 50.0, "resolved_count": null, "total_count": null, "cost_usd": 8.2182, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 8, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "claude-4-opus", "name": "Claude 4 Opus", "vendor": "Anthropic"}, "success_rate": 50.0, "resolved_count": null, "total_count": null, "cost_usd": null, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 9, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "gpt-5-medium", "name": "GPT-5 (medium)", "vendor": "OpenAI"}, "success_rate": 46.0, "resolved_count": null, "total_count": null, "cost_usd": 3.3251, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 10, "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": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "success_rate": 46.0, "resolved_count": null, "total_count": null, "cost_usd": 8.1618, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 11, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "o3", "name": "o3", "vendor": "openai"}, "success_rate": 46.0, "resolved_count": null, "total_count": null, "cost_usd": 9.8659, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "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": "claude-4-5-haiku-high", "name": "Claude 4.5 Haiku (high)", "vendor": "Anthropic"}, "success_rate": 44.0, "resolved_count": null, "total_count": null, "cost_usd": 1.3329, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 13, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "gpt-4-1", "name": "GPT-4.1", "vendor": "openai"}, "success_rate": 44.0, "resolved_count": null, "total_count": null, "cost_usd": 8.0337, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 14, "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": "claude-4-1-opus", "name": "Claude 4.1 Opus", "vendor": "Anthropic"}, "success_rate": 42.0, "resolved_count": null, "total_count": null, "cost_usd": 9.748, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 15, "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": "claude-4-5-sonnet-high", "name": "Claude 4.5 Sonnet (high)", "vendor": "Anthropic"}, "success_rate": 40.0, "resolved_count": null, "total_count": null, "cost_usd": 1.9586, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 16, "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": "claude-4-5-sonnet", "name": "Claude 4.5 Sonnet", "vendor": "Anthropic"}, "success_rate": 34.0, "resolved_count": null, "total_count": null, "cost_usd": 2.6161, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 17, "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": "claude-4-opus", "name": "Claude 4 Opus", "vendor": "Anthropic"}, "success_rate": 34.0, "resolved_count": null, "total_count": null, "cost_usd": 7.8039, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 18, "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": "claude-4-opus-high", "name": "Claude 4 Opus (high)", "vendor": "Anthropic"}, "success_rate": 30.0, "resolved_count": null, "total_count": null, "cost_usd": 8.2331, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 19, "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": "claude-3-7-sonnet", "name": "Claude 3.7 Sonnet", "vendor": "Anthropic"}, "success_rate": 26.0, "resolved_count": null, "total_count": null, "cost_usd": 2.3965, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 20, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "gemini-2-0-flash", "name": "Gemini 2.0 Flash", "vendor": "Google"}, "success_rate": 24.0, "resolved_count": null, "total_count": null, "cost_usd": 0.0963, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 21, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "deepseek-v3-march-2025", "name": "DeepSeek V3 (March 2025)", "vendor": "DeepSeek"}, "success_rate": 24.0, "resolved_count": null, "total_count": null, "cost_usd": 0.2402, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 22, "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": "claude-3-7-sonnet-high", "name": "Claude 3.7 Sonnet (high)", "vendor": "Anthropic"}, "success_rate": 24.0, "resolved_count": null, "total_count": null, "cost_usd": 1.4894, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 23, "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": "claude-4-5-haiku", "name": "Claude 4.5 Haiku", "vendor": "Anthropic"}, "success_rate": 24.0, "resolved_count": null, "total_count": null, "cost_usd": 3.0182, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 24, "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-5-medium", "name": "GPT-5 (medium)", "vendor": "OpenAI"}, "success_rate": 12.0, "resolved_count": null, "total_count": null, "cost_usd": 1.1751, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 25, "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": "deepseek-v3-march-2025", "name": "DeepSeek V3 (March 2025)", "vendor": "DeepSeek"}, "success_rate": 10.0, "resolved_count": null, "total_count": null, "cost_usd": 0.6157, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 26, "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": "deepseek-r1-january-2025", "name": "DeepSeek R1 (January 2025)", "vendor": "DeepSeek"}, "success_rate": 6.0, "resolved_count": null, "total_count": null, "cost_usd": 2.9941, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 27, "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": "o4-mini", "name": "o4 mini", "vendor": "openai"}, "success_rate": 2.0, "resolved_count": null, "total_count": null, "cost_usd": 0.6535, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 28, "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": "gemini-2-0-flash", "name": "Gemini 2.0 Flash", "vendor": "Google"}, "success_rate": 2.0, "resolved_count": null, "total_count": null, "cost_usd": 0.1496, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 29, "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": 2.0, "resolved_count": null, "total_count": null, "cost_usd": 1.0571, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 30, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "scaffold_code": null, "repo_url": "https://github.com/SWE-agent/SWE-agent", "homepage_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20240402_sweagent_claude3opus.png"}, "model": {"slug": "deepseek-r1-january-2025", "name": "DeepSeek R1 (January 2025)", "vendor": "DeepSeek"}, "success_rate": 0.0, "resolved_count": null, "total_count": null, "cost_usd": 0.0849, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}, {"rank": 31, "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": "o3", "name": "o3", "vendor": "openai"}, "success_rate": 0.0, "resolved_count": null, "total_count": null, "cost_usd": 11.9533, "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 SWE-bench Verified Mini (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/swebench_verified_mini", "observed_at": null}]}}