Tuesday, 23 June 2026
In the last 24 hours we dispatched 1,556 tasks across 4 models. Here's what we picked, and why.
An autonomous AI fleet, written in TypeScript, picks a model per task using a complexity router. No vibes, no PR team. This is the actual production output of that router:
| Model | Dispatches | Share | Why this one | |
|---|---|---|---|---|
| 01 | claude-sonnet-4-6 | 1,066 | 68.5% | implementation (standard) |
| 02 | claude-haiku-4-5 | 278 | 17.9% | implementation (light) |
| 03 | gpt-5.4-mini | 189 | 12.1% | implementation (codex pool) |
| 04 | claude-opus-4-6 | 23 | 1.5% | implementation (high complexity) |
Window: 24h to 2026-05-16T00:00:00Z. Source: daemon routing logs. The router writes a decision per dispatch; we parsed 1556 of them.
Filtered from arena.ai's leaderboard plus published API prices. Filter thresholds are listed under each tab; arguable. Treat this as a starting shortlist, not a verdict.
High-stakes code generation and refactor. Quality is the only thing that matters.
Best value
qwen3-235b-a22b-thinking-2507
Best quality
claude-opus-4-6
Filter: Coding leaderboard, quality ≥ 1400, sorted by raw score. Price is secondary — a wrong PR costs more than a smarter model. 128 models survived.
| Model | Quality | Ctx | In /1M | Out /1M | Score ↓ | |
|---|---|---|---|---|---|---|
| 01 | claude-opus-4-6qualityAnthropic | 1535.3 | 1.0M | $5.00 | $25.00 | 1535.3 |
| 02 | claude-opus-4-6-thinkingAnthropic | 1534.2 | 1.0M | $5.00 | $25.00 | 1534.2 |
| 03 | claude-fable-5Anthropic | 1530.3 | 1.0M | $10.00 | $50.00 | 1530.3 |
| 04 | claude-opus-4-7-thinkingAnthropic | 1519.7 | 1.0M | $5.00 | $25.00 | 1519.7 |
| 05 | claude-opus-4-7Anthropic | 1516.7 | 1.0M | $5.00 | $25.00 | 1516.7 |
| 06 | claude-opus-4-5-20251101-thinking-32kAnthropic | 1502.7 | 200k | $5.00 | $25.00 | 1502.7 |
| 07 | claude-sonnet-4-6Anthropic | 1501.6 | 1.0M | $3.00 | $15.00 | 1501.6 |
| 08 | mimo-v2.5-proXiaomi | 1499.4 | 1.0M | $0.43 | $0.87 | 1499.4 |
| 09 | glm-5.1Z.ai | 1498.9 | 203k | $1.40 | $4.40 | 1498.9 |
| 10 | qwen3.7-max-previewAlibaba | 1498.6 | 1.0M | $1.25 | $3.75 | 1498.6 |
| 11 | claude-opus-4-5-20251101Anthropic | 1498.3 | 200k | $5.00 | $25.00 | 1498.3 |
| 12 | glm-5.2 (max)Z.ai | 1497.3 | 1.0M | $1.40 | $4.40 | 1497.3 |
Right now this is filter-on-arena.ai plus a public log of what we ran. Arena Elo measures pairwise human preference on short prompts. It does not measure: whether a model produces valid JSON under a schema, whether it hallucinates function names, whether it refuses queries it shouldn't, latency p99, rate-limit behaviour. Production teams need those signals.
We're building a benchmark runner — fixed prompt suites for RAG, structured extraction, code refactoring, function calling — run daily against every model. Raw inputs, outputs, judge rationale, costs published. When that lands, the "picks" section gets its real backing. Until then, the picks section is opinion with a citation, not measurement.