{"model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic", "model_family": "claude-4-1-opus", "parameter_count_b": null, "context_length": 200000, "homepage_url": null, "hf_model_id": null, "openrouter_id": null, "card_summary": "Anthropic Claude 4.1 Opus (high) on SWE-bench Verified leaderboards.", "is_rl_checkpoint": false, "score_count": 12, "best_success_rate": 68.48, "input_price_per_million": null, "output_price_per_million": null}, "scores": [{"success_rate": 68.48, "resolved_count": null, "total_count": null, "cost_usd": 3.4075, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 165, "is_harnessrl_measured": false, "eval_protocol": "HAL GAIA (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/gaia", "observed_at": null, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-gaia", "name": "GAIA (HAL)"}}, {"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, "harness_version": null, "trial_count": 49, "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, "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-swe-bench-verified-mini", "name": "SWE-bench Verified Mini (HAL)"}}, {"success_rate": 52.0, "resolved_count": null, "total_count": null, "cost_usd": 2.9996, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 50, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-tau-bench-airline", "name": "TAU-bench Airline (HAL)"}}, {"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, "harness_version": null, "trial_count": 135, "is_harnessrl_measured": false, "eval_protocol": "HAL USACO (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/usaco", "observed_at": null, "harness": {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-usaco", "name": "USACO (HAL)"}}, {"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, "harness_version": null, "trial_count": 49, "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, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-swe-bench-verified-mini", "name": "SWE-bench Verified Mini (HAL)"}}, {"success_rate": 42.22, "resolved_count": null, "total_count": null, "cost_usd": 11.3322, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 45, "is_harnessrl_measured": false, "eval_protocol": "HAL CORE-Bench Hard (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/corebench_hard", "observed_at": null, "harness": {"slug": "core-agent", "name": "CORE-Agent", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (HAL)"}}, {"success_rate": 33.33, "resolved_count": null, "total_count": null, "cost_usd": 7.966, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 45, "is_harnessrl_measured": false, "eval_protocol": "HAL CORE-Bench Hard (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/corebench_hard", "observed_at": null, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (HAL)"}}, {"success_rate": 32.0, "resolved_count": null, "total_count": null, "cost_usd": 2.8056, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 50, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-tau-bench-airline", "name": "TAU-bench Airline (HAL)"}}, {"success_rate": 26.47, "resolved_count": null, "total_count": null, "cost_usd": 0.3309, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 102, "is_harnessrl_measured": false, "eval_protocol": "HAL ScienceAgentBench (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/scienceagentbench", "observed_at": null, "harness": {"slug": "sab-self-debug", "name": "SAB Self Debug", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-scienceagentbench", "name": "ScienceAgentBench (HAL)"}}, {"success_rate": 25.45, "resolved_count": null, "total_count": null, "cost_usd": 8.9312, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 165, "is_harnessrl_measured": false, "eval_protocol": "HAL GAIA (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/gaia", "observed_at": null, "harness": {"slug": "hf-open-deep-research", "name": "HF Open Deep Research", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-gaia", "name": "GAIA (HAL)"}}, {"success_rate": 13.75, "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, "harness_version": null, "trial_count": 30, "is_harnessrl_measured": false, "eval_protocol": "HAL AssistantBench (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/assistantbench", "observed_at": null, "harness": {"slug": "browser-use", "name": "Browser-Use", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-assistantbench", "name": "AssistantBench (HAL)"}}, {"success_rate": 6.92, "resolved_count": null, "total_count": null, "cost_usd": 6.8817, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 80, "is_harnessrl_measured": false, "eval_protocol": "HAL SciCode (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/scicode", "observed_at": null, "harness": {"slug": "scicode-tool-calling-agent", "name": "Scicode Tool Calling Agent", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-scicode", "name": "SciCode (HAL)"}}], "harnesses": [{"slug": "browser-use", "name": "Browser-Use", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, {"slug": "core-agent", "name": "CORE-Agent", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "repo_url": "https://github.com/princeton-pli/hal-harness", "openrouter_icon_url": null}, {"slug": "hf-open-deep-research", "name": "HF Open Deep Research", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, {"slug": "sab-self-debug", "name": "SAB Self Debug", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "repo_url": "https://github.com/SWE-agent/SWE-agent", "openrouter_icon_url": null}, {"slug": "scicode-tool-calling-agent", "name": "Scicode Tool Calling Agent", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, {"slug": "usaco-episodic-semantic", "name": "USACO Episodic + Semantic", "vendor": null, "repo_url": null, "openrouter_icon_url": null}]}