{"model": {"slug": "claude-4-5-opus-high", "name": "Claude 4.5 Opus (high)", "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 (high) on SWE-bench Verified leaderboards.", "is_rl_checkpoint": false, "score_count": 4, "best_success_rate": 76.8, "input_price_per_million": null, "output_price_per_million": null}, "scores": [{"success_rate": 76.8, "resolved_count": 384, "total_count": 500, "cost_usd": 0.753907997, "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/claude", "observed_at": "2026-02-17T00:00:00+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-high", "name": "Claude 4.5 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-bash-only", "name": "SWE-bench Verified (bash-only)"}}, {"success_rate": 76.8, "resolved_count": 384, "total_count": 500, "cost_usd": 0.753907997, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 33, "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": "2026-02-17T00:00:00+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-high", "name": "Claude 4.5 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 42.22, "resolved_count": null, "total_count": null, "cost_usd": 3.3924, "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-high", "name": "Claude 4.5 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (HAL)"}}, {"success_rate": 31.11, "resolved_count": null, "total_count": null, "cost_usd": 2.4973, "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-high", "name": "Claude 4.5 Opus (high)", "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": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "repo_url": "https://github.com/princeton-pli/hal-harness", "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}]}