{"model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic", "model_family": "claude-4-sonnet", "parameter_count_b": null, "context_length": 200000, "homepage_url": null, "hf_model_id": null, "openrouter_id": null, "card_summary": "Anthropic Claude 4 Sonnet \u2014 balanced frontier model for coding agents.", "is_rl_checkpoint": false, "score_count": 20, "best_success_rate": 92.75, "input_price_per_million": null, "output_price_per_million": null}, "scores": [{"success_rate": 92.75, "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/", "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-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-reliability-tau-bench-airline", "name": "HAL Reliability \u2014 \u03c4-bench Airline (clean)"}}, {"success_rate": 91.0, "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": 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-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-reliability-gaia", "name": "HAL Reliability \u2014 GAIA"}}, {"success_rate": 76.8, "resolved_count": 384, "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.epam.com/services/artificial-intelligence", "observed_at": "2025-08-04T00:00:00+00:00", "harness": {"slug": "epam-ai-run-developer-agent", "name": "EPAM AI Run", "vendor": "EPAM Systems, Inc.", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 74.6, "resolved_count": 373, "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://github.com/nimasteryang/Lingxi", "observed_at": "2025-07-20T00:00:00+00:00", "harness": {"slug": "lingxi-v1-5", "name": "Lingxi V1.5 Claude 4 Sonnet 20250514", "vendor": "Nimasteryang", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 72.4, "resolved_count": 362, "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.anthropic.com/", "observed_at": "2025-05-22T00:00:00+00:00", "harness": {"slug": "tools", "name": "Tools", "vendor": "Anthropic", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 70.8, "resolved_count": 354, "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://github.com/aorwall/moatless-tools", "observed_at": "2025-06-11T00:00:00+00:00", "harness": {"slug": "moatless-tools", "name": "Moatless Tools", "vendor": "Aorwall", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 70.4, "resolved_count": 352, "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://all-hands.dev/", "observed_at": "2025-05-24T00:00:00+00:00", "harness": {"slug": "openhands-sdk", "name": "OpenHands SDK", "vendor": "All Hands AI", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 66.6, "resolved_count": 333, "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://github.com/SWE-agent/SWE-agent", "observed_at": "2025-05-22T00:00:00+00:00", "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 64.93, "resolved_count": 325, "total_count": 500, "cost_usd": 0.37145316780000004, "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 bash-only (mini-SWE-agent)", "source": "swe_bench_site", "source_url": "https://www.anthropic.com/news/claude-4", "observed_at": "2025-07-26T00:00:00+00:00", "harness": {"slug": "mini-swe-agent", "name": "mini-SWE-agent", "vendor": "anthropic", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-bash-only", "name": "SWE-bench Verified (bash-only)"}}, {"success_rate": 64.93, "resolved_count": 325, "total_count": 500, "cost_usd": 0.37145316780000004, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 37, "is_harnessrl_measured": false, "eval_protocol": "SWE-bench Verified public submission", "source": "swe_bench_site", "source_url": "https://mini-swe-agent.com/latest/", "observed_at": "2025-07-26T00:00:00+00:00", "harness": {"slug": "mini-swe-agent", "name": "mini-SWE-agent", "vendor": "anthropic", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 64.21, "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-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-rebench", "name": "SWE-rebench"}}, {"success_rate": 60.33, "resolved_count": 181, "total_count": 300, "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 Lite public submission", "source": "swe_bench_site", "source_url": "https://github.com/ExpeRepair/ExpeRepair", "observed_at": "2025-06-25T00:00:00+00:00", "harness": {"slug": "experepair-v1-0", "name": "ExpeRepair V1.0", "vendor": "Experepair", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-lite", "name": "SWE-bench Lite"}}, {"success_rate": 59.43, "resolved_count": null, "total_count": 1055, "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": "LiveCodeBench generation pass@1 mean", "source": "livecodebench", "source_url": "https://livecodebench.github.io/leaderboard.html", "observed_at": null, "harness": {"slug": "livecodebench", "name": "Livecodebench", "vendor": "LiveCodeBench", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "livecodebench-generation", "name": "LiveCodeBench (code generation)"}}, {"success_rate": 58.33, "resolved_count": 175, "total_count": 300, "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 Lite public submission", "source": "swe_bench_site", "source_url": "https://github.com/GLEAM-Lab/KGCompass", "observed_at": "2025-09-06T00:00:00+00:00", "harness": {"slug": "kgcompass", "name": "KGCompass", "vendor": "Gleam Lab", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-lite", "name": "SWE-bench Lite"}}, {"success_rate": 56.67, "resolved_count": 170, "total_count": 300, "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 Lite public submission", "source": "swe_bench_site", "source_url": "https://github.com/SWE-agent/SWE-agent", "observed_at": "2025-05-26T00:00:00+00:00", "harness": {"slug": "swe-agent", "name": "SWE-agent", "vendor": "SWE-agent", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-lite", "name": "SWE-bench Lite"}}, {"success_rate": 46.67, "resolved_count": null, "total_count": null, "cost_usd": 1.4573, "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-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (HAL)"}}, {"success_rate": 40.0, "resolved_count": null, "total_count": null, "cost_usd": 5.2575, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 300, "is_harnessrl_measured": false, "eval_protocol": "HAL Online Mind2Web (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/online_mind2web", "observed_at": null, "harness": {"slug": "browser-use", "name": "Browser-Use", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-online-mind2web", "name": "Online Mind2Web (HAL)"}}, {"success_rate": 36.67, "resolved_count": null, "total_count": null, "cost_usd": 0.8206, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 300, "is_harnessrl_measured": false, "eval_protocol": "HAL Online Mind2Web (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/online_mind2web", "observed_at": null, "harness": {"slug": "seeact", "name": "SeeAct", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-online-mind2web", "name": "Online Mind2Web (HAL)"}}, {"success_rate": 35.59, "resolved_count": 184, "total_count": 517, "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 Multimodal public submission", "source": "swe_bench_site", "source_url": "https://refact.ai", "observed_at": "2025-06-11T00:00:00+00:00", "harness": {"slug": "refact-ai-agent", "name": "Refact Ai Agent", "vendor": "Refact.ai", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "claude-4-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-multimodal-site", "name": "SWE-bench Multimodal (site)"}}, {"success_rate": 28.89, "resolved_count": null, "total_count": null, "cost_usd": 1.1171, "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-sonnet", "name": "Claude 4 Sonnet", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (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": "claude-code", "name": "Claude Code", "vendor": "Anthropic", "repo_url": "https://github.com/anthropics/claude-code", "openrouter_icon_url": "https://claude.ai/favicon.ico"}, {"slug": "epam-ai-run-developer-agent", "name": "EPAM AI Run", "vendor": "EPAM Systems, Inc.", "repo_url": "https://github.com/epam/AI-RUN", "openrouter_icon_url": null}, {"slug": "experepair-v1-0", "name": "ExpeRepair V1.0", "vendor": "Experepair", "repo_url": "https://github.com/ExpeRepair/ExpeRepair", "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": "kgcompass", "name": "KGCompass", "vendor": "Gleam Lab", "repo_url": "https://github.com/GLEAM-Lab/KGCompass", "openrouter_icon_url": null}, {"slug": "lingxi-v1-5", "name": "Lingxi V1.5 Claude 4 Sonnet 20250514", "vendor": "Nimasteryang", "repo_url": "https://github.com/nimasteryang/Lingxi", "openrouter_icon_url": null}, {"slug": "livecodebench", "name": "Livecodebench", "vendor": "LiveCodeBench", "repo_url": "https://github.com/LiveCodeBench/LiveCodeBench", "openrouter_icon_url": null}, {"slug": "moatless-tools", "name": "Moatless Tools", "vendor": "Aorwall", "repo_url": "https://github.com/aorwall/moatless-tools", "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": "refact-ai-agent", "name": "Refact Ai Agent", "vendor": "Refact.ai", "repo_url": "https://github.com/smallcloudai/refact", "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": "seeact", "name": "SeeAct", "vendor": null, "repo_url": null, "openrouter_icon_url": null}, {"slug": "tools", "name": "Tools", "vendor": "Anthropic", "repo_url": "https://github.com/anthropics", "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}]}