{"model": {"slug": "deepseek-v3-1-august-2025", "name": "DeepSeek V3.1 (August 2025)", "vendor": "DeepSeek", "model_family": "deepseek-v3-1-august-2025", "parameter_count_b": null, "context_length": 128000, "homepage_url": null, "hf_model_id": null, "openrouter_id": null, "card_summary": "DeepSeek DeepSeek V3.1 (August 2025) on SWE-bench Verified leaderboards.", "is_rl_checkpoint": false, "score_count": 1, "best_success_rate": 20.0, "input_price_per_million": null, "output_price_per_million": null}, "scores": [{"success_rate": 20.0, "resolved_count": null, "total_count": null, "cost_usd": 0.2789, "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": "deepseek-v3-1-august-2025", "name": "DeepSeek V3.1 (August 2025)", "vendor": "DeepSeek"}, "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}]}