{"model": {"slug": "gemini-2-5-flash", "name": "Gemini 2.5 Flash", "vendor": "google", "model_family": "gemini-2-5-flash", "parameter_count_b": null, "context_length": 1048576, "homepage_url": null, "hf_model_id": null, "openrouter_id": "google/gemini-2.5-flash", "card_summary": "google Gemini 2.5 Flash on SWE-bench Verified leaderboards.", "is_rl_checkpoint": false, "score_count": 5, "best_success_rate": 85.5, "input_price_per_million": 0.3, "output_price_per_million": 2.5}, "scores": [{"success_rate": 85.5, "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": "HAL reliability composite (mean of 4 dimension aggregates)", "source": "hal_reliability_site", "source_url": "https://hal.cs.princeton.edu/reliability/benchmark/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": "gemini-2-5-flash", "name": "Gemini 2.5 Flash", "vendor": "google"}, "benchmark": {"slug": "hal-reliability-gaia", "name": "HAL Reliability \u2014 GAIA"}}, {"success_rate": 81.5, "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": "HAL reliability composite (mean of 4 dimension aggregates)", "source": "hal_reliability_site", "source_url": "https://hal.cs.princeton.edu/reliability/benchmark/taubench_airline_original/", "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-2-5-flash", "name": "Gemini 2.5 Flash", "vendor": "google"}, "benchmark": {"slug": "hal-reliability-tau-bench-airline-original", "name": "HAL Reliability \u2014 \u03c4-bench Airline (original)"}}, {"success_rate": 77.75, "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": "HAL reliability composite (mean of 4 dimension aggregates)", "source": "hal_reliability_site", "source_url": "https://hal.cs.princeton.edu/reliability/benchmark/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": "gemini-2-5-flash", "name": "Gemini 2.5 Flash", "vendor": "google"}, "benchmark": {"slug": "hal-reliability-tau-bench-airline", "name": "HAL Reliability \u2014 \u03c4-bench Airline (clean)"}}, {"success_rate": 76.21, "resolved_count": null, "total_count": 1055, "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": "LiveCodeBench generation pass@1 mean", "source": "livecodebench", "source_url": "https://livecodebench.github.io/leaderboard.html", "observed_at": null, "harness": {"slug": "livecodebench", "name": "Livecodebench", "vendor": "LiveCodeBench", "repo_url": null, "openrouter_icon_url": null}, "model": {"slug": "gemini-2-5-flash", "name": "Gemini 2.5 Flash", "vendor": "google"}, "benchmark": {"slug": "livecodebench-generation", "name": "LiveCodeBench (code generation)"}}, {"success_rate": 35.71, "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-2-5-flash", "name": "Gemini 2.5 Flash", "vendor": "google"}, "benchmark": {"slug": "swe-rebench", "name": "SWE-rebench"}}], "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}, {"slug": "livecodebench", "name": "Livecodebench", "vendor": "LiveCodeBench", "repo_url": "https://github.com/LiveCodeBench/LiveCodeBench", "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}]}