{"model": {"slug": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google", "model_family": "gemini-3-pro-preview", "parameter_count_b": null, "context_length": 1000000, "homepage_url": null, "hf_model_id": null, "openrouter_id": null, "card_summary": "Google Gemini 3 Pro Preview on SWE-bench Verified leaderboards.", "is_rl_checkpoint": false, "score_count": 11, "best_success_rate": 77.4, "input_price_per_million": null, "output_price_per_million": null}, "scores": [{"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": null, "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"}}, {"success_rate": 74.2, "resolved_count": 371, "total_count": 500, "cost_usd": 0.45995751895000003, "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://ai.google.dev/gemini-api/docs/models", "observed_at": "2025-11-18T00:00:00+00:00", "harness": {"slug": "mini-swe-agent", "name": "mini-SWE-agent", "vendor": "anthropic", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "benchmark": {"slug": "swe-bench-bash-only", "name": "SWE-bench Verified (bash-only)"}}, {"success_rate": 74.2, "resolved_count": 371, "total_count": 500, "cost_usd": 0.45995751895000003, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "harness_version": null, "trial_count": 40, "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": "2025-11-18T00:00:00+00:00", "harness": {"slug": "mini-swe-agent", "name": "mini-SWE-agent", "vendor": "anthropic", "repo_url": null, "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"}}, {"success_rate": 67.27, "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": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "benchmark": {"slug": "swe-rebench", "name": "SWE-rebench"}}, {"success_rate": 64.44, "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": "MLE-bench All split any_medal_percentage", "source": "mle_bench_site", "source_url": "https://github.com/openai/mle-bench", "observed_at": "2026-02-23T00:00:00+00:00", "harness": {"slug": "famou-agent-2-0", "name": "Famou Agent 2.0", "vendor": "Baidubce", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "benchmark": {"slug": "mle-bench", "name": "MLE-bench"}}, {"success_rate": 61.33, "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": "MLE-bench All split any_medal_percentage", "source": "mle_bench_site", "source_url": "https://github.com/openai/mle-bench", "observed_at": "2026-02-14T00:00:00+00:00", "harness": {"slug": "mlevolve", "name": "MLEvolve", "vendor": "Internscience", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "benchmark": {"slug": "mle-bench", "name": "MLE-bench"}}, {"success_rate": 56.0, "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": "MLE-bench All split any_medal_percentage", "source": "mle_bench_site", "source_url": "https://github.com/openai/mle-bench", "observed_at": "2026-01-25T00:00:00+00:00", "harness": {"slug": "cair-mars", "name": "Cair Mars", "vendor": "Google", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "benchmark": {"slug": "mle-bench", "name": "MLE-bench"}}, {"success_rate": 52.0, "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": "MLE-bench All split any_medal_percentage", "source": "mle_bench_site", "source_url": "https://github.com/openai/mle-bench", "observed_at": "2026-01-05T00:00:00+00:00", "harness": {"slug": "pievolve", "name": "PiEvolve", "vendor": "Fractalairesearchlabs", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "benchmark": {"slug": "mle-bench", "name": "MLE-bench"}}, {"success_rate": 50.67, "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": "MLE-bench All split any_medal_percentage", "source": "mle_bench_site", "source_url": "https://github.com/openai/mle-bench", "observed_at": "2025-12-07T00:00:00+00:00", "harness": {"slug": "leeroo", "name": "Leeroo", "vendor": "Leeroo Ai", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "benchmark": {"slug": "mle-bench", "name": "MLE-bench"}}, {"success_rate": 40.0, "resolved_count": null, "total_count": null, "cost_usd": 1.9244, "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": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "benchmark": {"slug": "hal-corebench-hard", "name": "CORE-Bench Hard (HAL)"}}, {"success_rate": 35.56, "resolved_count": null, "total_count": null, "cost_usd": 2.2504, "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": "gemini-3-pro-preview", "name": "Gemini 3 Pro Preview", "vendor": "Google"}, "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": "cair-mars", "name": "Cair Mars", "vendor": "Google", "repo_url": null, "openrouter_icon_url": null}, {"slug": "famou-agent-2-0", "name": "Famou Agent 2.0", "vendor": "Baidubce", "repo_url": "https://github.com/baidubce/FM-Agent", "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": "leeroo", "name": "Leeroo", "vendor": "Leeroo Ai", "repo_url": "https://github.com/Leeroo-AI/kapso", "openrouter_icon_url": null}, {"slug": "live-swe-agent", "name": "Live SWE Agent", "vendor": "OpenAutoCoder", "repo_url": "https://github.com/OpenAutoCoder/live-swe-agent", "openrouter_icon_url": null}, {"slug": "mlevolve", "name": "MLEvolve", "vendor": "Internscience", "repo_url": "https://github.com/InternScience/MLEvolve", "openrouter_icon_url": null}, {"slug": "pievolve", "name": "PiEvolve", "vendor": "Fractalairesearchlabs", "repo_url": "https://github.com/FractalAIResearchLabs/PiEvolve", "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}]}