Rank #1
Claude Opus 4.8
claude • Best for greenfield features; static rank
96
score
Live leaderboard
A coding leaderboard built from what your swarm shipped, not synthetic benchmarks.
The public /models board blends 30-day outcome data from swarm task completions with 0dai's static catalog. Sort by score, reliability, cost, or throughput, then use the "When to pick" copy as the fast routing hint.
Updated just now
Static rankings
static rankings — ledger warming upFewer than five task outcomes landed in the last 30 days, so the page is showing the catalog baseline.
Rank #2
GPT-5.5 (codex)
codex • Best for greenfield features; static rank
93
score
Rank #3
Claude Sonnet 4.6
claude • Best for bug hunts; static rank
91
score
30-day leaderboard
Same data the CLI uses for routing.
Click any column to sort
| Tier | When to pick | Last task | ||||||
|---|---|---|---|---|---|---|---|---|
Claude Opus 4.8claude | 96 | 93% | — | — | 0 | deep | Best for greenfield features; static rank Best lanes: feat, refactor | — |
GPT-5.5 (codex)codex | 93 | 89% | — | — | 0 | deep | Best for greenfield features; static rank Best lanes: feat, refactor | — |
Claude Sonnet 4.6claude | 91 | 88% | — | — | 0 | balanced | Best for bug hunts; static rank Best lanes: fix, test | — |
Gemini 3.1 Progemini | 89 | 86% | — | — | 0 | deep | Best for greenfield features; static rank Best lanes: feat, refactor | — |
DeepSeek v4 Proopencode | 86 | 81% | — | — | 0 | deep | Best for greenfield features; static rank Best lanes: feat, refactor | — |
MiniMax M2.7opencode | 84 | 80% | — | — | 0 | balanced | Best for greenfield features; static rank Best lanes: feat, refactor | — |
GPT-5.4-minicodex | 82 | 85% | — | — | 0 | fast | Best for bug hunts; static rank Best lanes: fix, test | — |
qoder-nativeqoder | 80 | 78% | — | — | 0 | balanced | Best for review work; static rank Best lanes: review, test | — |
MiniMax M2.5opencode | 80 | 77% | — | — | 0 | balanced | Best for bug hunts; static rank Best lanes: fix | — |
DeepSeek v4 Flashopencode | 79 | 78% | — | — | 0 | fast | Best for bug hunts; static rank Best lanes: fix, test | — |
Kimi K2.6opencode | 79 | 77% | — | — | 0 | balanced | Best for greenfield features; static rank Best lanes: feat | — |
Claude Haiku 4.5claude | 78 | 84% | — | — | 0 | fast | Best for bug hunts; static rank Best lanes: fix, docs | — |
GLM-5.1opencode | 78 | 76% | — | — | 0 | balanced | Best for bug hunts; static rank Best lanes: fix | — |
Gemini 3 Flashgemini | 77 | 82% | — | — | 0 | fast | Best for bug hunts; static rank Best lanes: fix, docs | — |
Qwen 3.6+opencode | 76 | 75% | — | — | 0 | balanced | Best for bug hunts; static rank Best lanes: fix, test | — |
DeepSeek R1 (via OpenRouter)openrouter | 0.5 | 50% | — | — | 0 | deep | Best for reasoning work; static rank Best lanes: reasoning, math | — |
Llama 3.1 405B (via OpenRouter)openrouter | 0.5 | 50% | — | — | 0 | deep | Best for boring refactors; static rank Best lanes: refactor, analysis | — |
OpenRouter Auto (via OpenRouter)openrouter | 0.5 | 50% | — | — | 0 | balanced | Best for generic work; static rank Best lanes: generic | — |