diff --git a/apps/web/src/pages/leaderboard.astro b/apps/web/src/pages/leaderboard.astro
new file mode 100644
index 000000000..d56908796
--- /dev/null
+++ b/apps/web/src/pages/leaderboard.astro
@@ -0,0 +1,632 @@
+---
+/**
+ * AgentV Leaderboard — SWE-bench Lite
+ *
+ * Static page that reads benchmark result JSON files at build time
+ * and renders a sortable table + Pareto frontier chart.
+ *
+ * Data source: /benchmarks/swe-bench-lite/results/*.json
+ * Route: /leaderboard
+ */
+
+// Read result files at build time
+import { readFileSync, readdirSync } from 'node:fs';
+import { join } from 'node:path';
+
+interface ResultData {
+ model: string;
+ provider: string;
+ model_type: string;
+ date: string;
+ agent: string;
+ agent_version: string;
+ dataset: string;
+ total_instances: number;
+ resolved_instances: number;
+ resolution_rate: number;
+ avg_cost_usd: number;
+ avg_cost_per_fix_usd: number;
+ avg_duration_ms: number;
+ avg_tool_calls: number;
+}
+
+const resultsDir = join(process.cwd(), '../../benchmarks/swe-bench-lite/results');
+let results: ResultData[] = [];
+
+// Sanitize string for use in CSS class names (alphanumeric + hyphens only)
+function safeCssClass(s: string): string {
+ return s.replace(/[^a-z0-9-]/gi, '').toLowerCase();
+}
+
+try {
+ const files = readdirSync(resultsDir).filter((f) => f.endsWith('.json'));
+ results = files.map((f) => {
+ const data = JSON.parse(readFileSync(join(resultsDir, f), 'utf8'));
+ return data as ResultData;
+ });
+ // Sort by resolution rate descending
+ results.sort((a, b) => b.resolution_rate - a.resolution_rate);
+} catch {
+ // Results dir may not exist in all environments
+}
+
+// Provider colors for chart
+const providerColors: Record
= {
+ anthropic: '#06b6d4',
+ openai: '#10b981',
+ google: '#f59e0b',
+ deepseek: '#8b5cf6',
+ meta: '#ef4444',
+};
+
+// Compute Pareto frontier
+function computeParetoFrontier(data: ResultData[]): ResultData[] {
+ const sorted = [...data].sort((a, b) => a.avg_cost_usd - b.avg_cost_usd);
+ const frontier: ResultData[] = [];
+ let maxRate = -1;
+ for (const d of sorted) {
+ if (d.resolution_rate > maxRate) {
+ frontier.push(d);
+ maxRate = d.resolution_rate;
+ }
+ }
+ return frontier;
+}
+
+const frontier = computeParetoFrontier(results);
+const frontierSet = new Set(frontier.map((f) => f.model));
+const providers = [...new Set(results.map((r) => r.provider))].sort();
+---
+
+
+
+
+
+
+ Leaderboard — AgentV SWE-bench Lite
+
+
+
+
+
+
+
+
+
+
+
+
+
+ AgentV Leaderboard — SWE-bench Lite
+ The multi-dimensional agent benchmark. Same SWE-bench tasks, richer metrics — cost efficiency, tool usage, and Pareto-optimal rankings.
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ | # ▲ |
+ Model ▲ |
+ Provider ▲ |
+ % Resolved ▲ |
+ Avg $ ▲ |
+ $/Fix ▲ |
+ Tools ▲ |
+ Latency ▲ |
+ Date ▲ |
+
+
+
+ {results.map((r, i) => {
+ const costClass = r.avg_cost_per_fix_usd < 0.5 ? 'good' : r.avg_cost_per_fix_usd < 0.8 ? 'mid' : 'bad';
+ const isFrontier = frontierSet.has(r.model);
+ return (
+
+ | {i + 1} |
+
+ {r.model}
+ {isFrontier && }
+ |
+ {r.provider} |
+ {(r.resolution_rate * 100).toFixed(1)}% |
+ ${r.avg_cost_usd.toFixed(2)} |
+ ${r.avg_cost_per_fix_usd.toFixed(2)} |
+ {r.avg_tool_calls.toFixed(1)} |
+ {(r.avg_duration_ms / 1000).toFixed(0)}s |
+ {r.date} |
+
+ );
+ })}
+
+
+
+
+
+
+ Pareto Frontier — Score vs Cost
+ Models on the frontier line achieve the best resolution rate for their cost. Closer to top-left is better.
+
+
+
+
+
+
+
+
+ Run it yourself
+
+
$ git clone https://github.com/EntityProcess/agentv
+
$ cd agentv/benchmarks/swe-bench-lite
+
$ bun run setup.ts
+
$ agentv eval ./evals/ --target claude
+
+
+
+ Submit your results →
+
+
+
+
+
+