$

llm-rate

Tuesday, 23 June 2026

In the last 24 hours we dispatched 1,556 tasks across 4 models. Here's what we picked, and why.

What we ran

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:

ModelDispatchesShareWhy 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.

If you don't have a router, here are the picks per common task

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.

Complex multi-step problems. Where to send the gnarly stuff.

Best value

qwen3-235b-a22b-thinking-2507

Alibaba · quality 1412.2 · $0.10/M blended

Best quality

gemini-3.5-flash

Google · quality 1519.7 · $6.75/M blended

Filter: Math leaderboard, quality ≥ 1400. Sorted by quality. Use the cheaper end of this list as a router when you're sure the task is hard. 113 models survived.

Model Quality Ctx In /1M Out /1M Score ↓
01 gemini-3.5-flashqualityGoogle 1519.7 1.0M $1.50 $9.00 1519.7
02 claude-opus-4-6-thinkingAnthropic 1514.6 1.0M $5.00 $25.00 1514.6
03 claude-fable-5Anthropic 1513.0 1.0M $10.00 $50.00 1513.0
04 claude-opus-4-6Anthropic 1508.1 1.0M $5.00 $25.00 1508.1
05 claude-opus-4-7-thinkingAnthropic 1497.7 1.0M $5.00 $25.00 1497.7
06 gpt-5.4-highOpenAI 1497.2 1.1M $2.50 $15.00 1497.2
07 qwen3.7-max-previewAlibaba 1495.3 1.0M $1.25 $3.75 1495.3
08 claude-opus-4-7Anthropic 1493.5 1.0M $5.00 $25.00 1493.5
09 gemini-3.1-pro-previewGoogle 1491.4 1.0M $2.00 $12.00 1491.4
10 claude-opus-4-8-thinkingAnthropic 1488.5 1.0M $5.00 $25.00 1488.5
11 claude-opus-4-8Anthropic 1487.6 1.0M $5.00 $25.00 1487.6
12 mimo-v2.5-proXiaomi 1484.2 1.0M $0.43 $0.87 1484.2

What this is, and isn't

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.