# Worker-tier routing

> Route architecture to a frontier planner and implementation to a cheap executor - a DeepSeek Flash-class executor at roughly 1/20-1/40 of Opus output cost needs a complete plan.

URL: https://bartlomiejkrupa.dev/notes/worker-tier-routing/
Category: concept | Updated: 2026-10-07
Summary: Route architecture to a frontier planner and implementation to a cheap executor - a DeepSeek Flash-class executor at roughly 1/20-1/40 of Opus output cost needs a complete plan.

**Worker-tier routing** sends planning to a strong model and execution to a cheaper tier. Quality lives in the plan; the executor runs the spec.

| Tier | $/M in/out (approx.) | Role |
| --- | --- | --- |
| Fable 5 | $10 / $50 | Rare hardest-step escalation ([refusals](/notes/refusal-as-api-contract)) |
| Opus 5 | $5 / $25 | Long-horizon orchestration |
| Sonnet 5 / 5.5 | $2 / $10 | Default worker |
| Haiku 4.5 | $1 / $5 | Parallel sub-agents |
| DeepSeek V4.1 Flash | $0.15-0.30 / $0.60-1.20 (off-peak / peak) | Cheap executor (complete spec required) |

DeepSeek V4 Flash scored 79.0 on SWE-bench Verified (DeepSeek's model card) against a reported 96.0 for Opus 5, and 60/100 on a real agentic build - strong at executing a complete plan, weak when inventing architecture. One Flash output token costs ~1/42nd of an Opus 5 token off-peak and ~1/21st at peak (V4.1 pricing since 2026-09-10). Practitioners calibrate it as Sonnet-class for clear specs, not Opus-class for vague ones.

## Conditions that make the split work

1. The plan is genuinely complete - every ambiguity becomes an invention.
2. The prompt prefix stays stable - cache hits are where the economics live.
3. Still [verify before done](/notes/verify-before-done); a cheap worker is not an excuse to skip checks.
4. [Effort vs fast mode](/notes/effort-vs-fast-mode) are separate dials on Claude tiers - do not confuse them with model choice.

Evidence: [DeepSeek V4 Flash is a worker tier, not an Opus killer](/articles/deepseek-v4-flash). Stack map: [Why most agents default to the wrong Claude tier](/articles/claude-model-tier-comparison).

Sources:
- Pricing: https://platform.claude.com/docs/en/about-claude/pricing
- Models & Pricing: https://api-docs.deepseek.com/quick_start/pricing
- DeepSeek-V4-Flash model card: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash