{"harness": {"slug": "live-swe-agent", "name": "Live SWE Agent", "vendor": "OpenAutoCoder", "description": "live-SWE-agent is a community fork/scaffold from OpenAutoCoder built on the SWE-agent lineage with configurations tuned for high SWE-bench Verified numbers. Public leaderboard rows often pair it with Claude Opus-class models.\n\nIt is a research/community harness (`status` community), not a HarnessRL validated scaffold. Scores reflect the fork's prompts, tool set, and model pairing at submission time.\n\nCatalog data comes from `swe_bench_site` ingest. For apples-to-apples model comparisons, contrast with mini-SWE-agent bash-only rows on the same benchmark rather than assuming identical tooling.", "card_summary": "OpenAutoCoder's live SWE-agent fork \u2014 strong public Verified results with frontier Claude pairings.", "homepage_url": "https://github.com/OpenAutoCoder/live-swe-agent", "repo_url": "https://github.com/OpenAutoCoder/live-swe-agent", "docs_url": null, "scaffold_code": null, "harbor_agent_name": null, "api_protocol": "openai", "status": "validated", "supports_rl": true, "score_count": 2, "best_success_rate": 79.2, "github_stars": 450, "popularity_tier": null, "hrl_score": 30.0, "hrl_rank": 31, "catalog_token_volume": null, "openrouter_icon_url": null, "leaderboard_icon_url": "https://www.swebench.com/img/logos/20251120_livesweagent_gemini-3-pro-preview.png", "aa_coding_index": null, "aa_mean_cost_usd": null, "aa_mean_total_tokens": null}, "scores": [{"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://github.com/OpenAutoCoder/live-swe-agent", "observed_at": "2025-12-15T00:00:00+00:00", "harness": {"slug": "live-swe-agent", "name": "Live SWE Agent", "vendor": "OpenAutoCoder", "repo_url": "https://github.com/OpenAutoCoder/live-swe-agent", "openrouter_icon_url": null}, "model": {"slug": "claude-4-5-opus-medium", "name": "Claude 4.5 Opus (medium)", "vendor": "Anthropic"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}, {"success_rate": 77.4, "resolved_count": 387, "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/OpenAutoCoder/live-swe-agent", "observed_at": "2025-11-20T00:00:00+00:00", "harness": {"slug": "live-swe-agent", "name": "Live SWE Agent", "vendor": "OpenAutoCoder", "repo_url": "https://github.com/OpenAutoCoder/live-swe-agent", "openrouter_icon_url": null}, "model": {"slug": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "benchmark": {"slug": "swe-bench-verified", "name": "SWE-bench Verified"}}], "benchmarks": [{"slug": "swe-bench-verified", "name": "SWE-bench Verified", "category": "eval"}]}