{"model": {"slug": "claude-4-opus-high", "name": "Claude 4 Opus (high)", "vendor": "Anthropic", "model_family": "claude-4-opus", "parameter_count_b": null, "context_length": 200000, "homepage_url": null, "hf_model_id": null, "openrouter_id": null, "card_summary": "Anthropic Claude 4 Opus (high) on SWE-bench Verified leaderboards.", "is_rl_checkpoint": false, "score_count": 3, "best_success_rate": 64.85, "input_price_per_million": null, "output_price_per_million": null}, "scores": [{"success_rate": 64.85, "resolved_count": null, "total_count": null, "cost_usd": 4.0357, "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-opus-high", "name": "Claude 4 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-gaia", "name": "GAIA (HAL)"}}, {"success_rate": 44.0, "resolved_count": null, "total_count": null, "cost_usd": 3.0058, "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-opus-high", "name": "Claude 4 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-tau-bench-airline", "name": "TAU-bench Airline (HAL)"}}, {"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, "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-opus-high", "name": "Claude 4 Opus (high)", "vendor": "Anthropic"}, "benchmark": {"slug": "hal-swe-bench-verified-mini", "name": "SWE-bench Verified Mini (HAL)"}}], "harnesses": [{"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "repo_url": "https://github.com/princeton-pli/hal-harness", "openrouter_icon_url": null}]}