{"model": {"slug": "gpt-oss-120b-august-2025", "name": "GPT OSS 120B (August 2025)", "vendor": "OpenAI", "model_family": "gpt-oss-120b-august-2025", "parameter_count_b": 120.0, "context_length": 400000, "homepage_url": null, "hf_model_id": null, "openrouter_id": null, "card_summary": "OpenAI GPT OSS 120B (August 2025) on SWE-bench Verified leaderboards.", "is_rl_checkpoint": false, "score_count": 2, "best_success_rate": 11.11, "input_price_per_million": null, "output_price_per_million": null}, "scores": [{"success_rate": 11.11, "resolved_count": null, "total_count": null, "cost_usd": 0.0936, "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": "gpt-oss-120b-august-2025", "name": "GPT OSS 120B (August 2025)", "vendor": "OpenAI"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (HAL)"}}, {"success_rate": 8.89, "resolved_count": null, "total_count": null, "cost_usd": 0.062, "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": "gpt-oss-120b-august-2025", "name": "GPT OSS 120B (August 2025)", "vendor": "OpenAI"}, "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}]}