{"model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic", "model_family": "claude-4-5-opus", "parameter_count_b": null, "context_length": 200000, "homepage_url": null, "hf_model_id": null, "openrouter_id": null, "card_summary": "Anthropic Claude 4.5 Opus on SWE-bench Verified leaderboards.", "is_rl_checkpoint": false, "score_count": 15, "best_success_rate": 96.25, "input_price_per_million": null, "output_price_per_million": null}, "scores": [{"success_rate": 96.25, "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": null, "is_harnessrl_measured": false, "eval_protocol": "HAL reliability composite (mean of 4 dimension aggregates)", "source": "hal_reliability_site", "source_url": "https://hal.cs.princeton.edu/reliability/benchmark/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-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-reliability-gaia", "name": "HAL Reliability \u2014 GAIA"}}, {"success_rate": 93.5, "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": null, "is_harnessrl_measured": false, "eval_protocol": "HAL reliability composite (mean of 4 dimension aggregates)", "source": "hal_reliability_site", "source_url": "https://hal.cs.princeton.edu/reliability/benchmark/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-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-reliability-tau-bench-airline", "name": "HAL Reliability \u2014 \u03c4-bench Airline (clean)"}}, {"success_rate": 87.75, "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": null, "is_harnessrl_measured": false, "eval_protocol": "HAL reliability composite (mean of 4 dimension aggregates)", "source": "hal_reliability_site", "source_url": "https://hal.cs.princeton.edu/reliability/benchmark/taubench_airline_original/", "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-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-reliability-tau-bench-airline-original", "name": "HAL Reliability \u2014 \u03c4-bench Airline (original)"}}, {"success_rate": 79.2, "resolved_count": 396, "total_count": 500, "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": null, "is_harnessrl_measured": false, "eval_protocol": "SWE-bench Verified public submission", "source": "swe_bench_site", "source_url": "https://www.sonarsource.com", "observed_at": "2025-12-05T00:00:00+00:00", "harness": {"slug": "sonar-foundation-agent", "name": "Sonar Foundation Agent", "vendor": "Sonar", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 78.5, "resolved_count": null, "total_count": null, "cost_usd": 1.38, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": null, "is_harnessrl_measured": false, "eval_protocol": "OpenHands SDK v1.8.3", "source": "openhands_index", "source_url": "https://index.openhands.dev/", "observed_at": "2026-07-31T11:38:58.577710", "harness": {"slug": "openhands-sdk", "name": "OpenHands SDK", "vendor": "All Hands AI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "swt-bench", "name": "SWT-Bench"}}, {"success_rate": 77.78, "resolved_count": null, "total_count": null, "cost_usd": 1.9369, "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": "claude-code", "name": "Claude Code", "vendor": "Anthropic", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (HAL)"}}, {"success_rate": 76.6, "resolved_count": null, "total_count": null, "cost_usd": 1.82, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": null, "is_harnessrl_measured": false, "eval_protocol": "OpenHands SDK v1.8.3", "source": "openhands_index", "source_url": "https://index.openhands.dev/", "observed_at": "2026-07-31T11:38:58.577710", "harness": {"slug": "openhands-sdk", "name": "OpenHands SDK", "vendor": "All Hands AI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "openhands-swe-bench", "name": "SWE-bench (OpenHands Index)"}}, {"success_rate": 70.7, "resolved_count": 212, "total_count": 300, "cost_usd": 0.8335529758333333, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": null, "is_harnessrl_measured": false, "eval_protocol": "SWE-bench Multilingual (mini-SWE-agent)", "source": "swe_bench_site", "source_url": "https://www.anthropic.com/news/claude-opus-4-5", "observed_at": "2026-02-13T00:00:00+00:00", "harness": {"slug": "mini-swe-agent", "name": "mini-SWE-agent", "vendor": "anthropic", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-multilingual", "name": "SWE-bench Multilingual"}}, {"success_rate": 69.1, "resolved_count": null, "total_count": null, "cost_usd": 0.55, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": null, "is_harnessrl_measured": false, "eval_protocol": "OpenHands SDK v1.8.3", "source": "openhands_index", "source_url": "https://index.openhands.dev/", "observed_at": "2026-07-31T11:38:58.577710", "harness": {"slug": "openhands-sdk", "name": "OpenHands SDK", "vendor": "All Hands AI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "gaia", "name": "GAIA"}}, {"success_rate": 68.89, "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": null, "is_harnessrl_measured": false, "eval_protocol": "SWE-rebench resolved rate (best rolling window)", "source": "swe_rebench_site", "source_url": "https://swe-rebench.com/", "observed_at": null, "harness": {"slug": "mini-swe-agent", "name": "mini-SWE-agent", "vendor": "anthropic", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-rebench", "name": "SWE-rebench"}}, {"success_rate": 52.62, "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": null, "is_harnessrl_measured": false, "eval_protocol": "SWE-bench Test public submission", "source": "swe_bench_site", "source_url": "https://www.sonarsource.com", "observed_at": "2025-12-19T00:00:00+00:00", "harness": {"slug": "sonar-foundation-agent", "name": "Sonar Foundation Agent", "vendor": "Sonar", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-test", "name": "SWE-bench Test"}}, {"success_rate": 42.22, "resolved_count": null, "total_count": null, "cost_usd": 3.7553, "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-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (HAL)"}}, {"success_rate": 41.2, "resolved_count": null, "total_count": null, "cost_usd": 2.54, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": null, "is_harnessrl_measured": false, "eval_protocol": "OpenHands SDK v1.8.3", "source": "openhands_index", "source_url": "https://index.openhands.dev/", "observed_at": "2026-07-31T11:38:58.577710", "harness": {"slug": "openhands-sdk", "name": "OpenHands SDK", "vendor": "All Hands AI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-multimodal", "name": "SWE-bench Multimodal (OpenHands Index)"}}, {"success_rate": 37.5, "resolved_count": null, "total_count": null, "cost_usd": 9.11, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": null, "is_harnessrl_measured": false, "eval_protocol": "OpenHands SDK v1.8.3", "source": "openhands_index", "source_url": "https://laminar.sh/shared/evals/d15f9cba-5c46-4413-b71d-f7b5d15e6efb", "observed_at": "2026-07-31T11:38:58.577710", "harness": {"slug": "openhands-sdk", "name": "OpenHands SDK", "vendor": "All Hands AI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "commit0", "name": "Commit0"}}, {"success_rate": 33.33, "resolved_count": null, "total_count": null, "cost_usd": 2.8313, "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-4-5-opus", "name": "Claude 4.5 Opus", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (HAL)"}}], "harnesses": [{"slug": "core-agent", "name": "CORE-Agent", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, {"slug": "claude-code", "name": "Claude Code", "vendor": "Anthropic", "repo_url": "https://github.com/anthropics/claude-code", "openrouter_icon_url": "https://claude.ai/favicon.ico"}, {"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": "openhands-sdk", "name": "OpenHands SDK", "vendor": "All Hands AI", "repo_url": "https://github.com/All-Hands-AI/OpenHands", "openrouter_icon_url": "https://www.all-hands.dev/favicon.ico"}, {"slug": "sonar-foundation-agent", "name": "Sonar Foundation Agent", "vendor": "Sonar", "repo_url": null, "openrouter_icon_url": null}, {"slug": "mini-swe-agent", "name": "mini-SWE-agent", "vendor": "anthropic", "repo_url": "https://github.com/SWE-agent/mini-swe-agent", "openrouter_icon_url": null}]}