---
schema_version: "newruntime-agent-readable-v0.2"
type: "post"
stable_id: "post:coding-agents-multi-model-harness"
slug: "coding-agents-multi-model-harness"
title: "The Coding Harness Is Becoming Independent From the Model"
description: "Practitioners are routing different models through coding-agent workflows, while production systems increasingly choose model and effort per role instead of per product."
retrieval_nugget: "Practitioners are routing different models through coding-agent workflows, while production systems increasingly choose model and effort per role instead of per product. A practitioner setup combining GPT-5.6 and Kimi K3 with a Claude Code-oriented workflow is a signal that the coding harness and the inference model are becoming separable layers."
status: "published"
published_at: "2026-07-24"
updated_at: "2026-07-24"
record_date: "2026-07-24"
date_kind: "published_at"
topics: ["coding-agents","model-routing","agent-harnesses"]
source_urls: ["https://x.com/theo/status/2078217355780624864","https://x.com/thsottiaux/status/2076119366647894371","https://docs.anthropic.com/en/docs/claude-code/llm-gateway","https://learn.chatgpt.com/docs/agent-configuration/subagents","https://www.uber.com/blog/ureview/"]
visuals: [{"id":"multi-model-harness-routing","kind":"editorial-diagram","role":"hero","src":"https://newruntime.com/images/posts/multi-model-harness-routing.webp","alt":"A portable coding harness routes work to four swappable model roles, then converges their outputs through one shared contract and deterministic validator.","caption":"Model mixing becomes useful when role boundaries, schemas, provenance, tests, and success conditions remain portable across providers.","credit":"New Runtime synthesis from gateway, subagent, and production-review patterns","source_url":"https://code.claude.com/docs/en/llm-gateway","generated_with":"gemini-3.1-flash-image","width":1600,"height":900,"legend":[{"label":"Portable harness","description":"Repository context, tools, permissions, planning, diffs, and verification remain the stable operating layer."},{"label":"Role routing","description":"Use distinct lanes for exploration, architecture, independent critique, and visual interface work."},{"label":"Swappable models","description":"Change a model only where the role has a measurable success condition and supported interface."},{"label":"Shared contract","description":"Make outputs converge through explicit schemas and provenance instead of incompatible opinions."},{"label":"Deterministic validation","description":"Tests and state-changing code decide whether the combined result becomes a verified artifact."}]}]
routes: {"html":"https://newruntime.com/posts/coding-agents-multi-model-harness/","markdown":"https://newruntime.com/posts/coding-agents-multi-model-harness.md","json":"https://newruntime.com/posts/coding-agents-multi-model-harness.json"}
source_format: "markdown"
---

# The Coding Harness Is Becoming Independent From the Model

## Retrieval answer

Practitioners are routing different models through coding-agent workflows, while production systems increasingly choose model and effort per role instead of per product. A practitioner setup combining GPT-5.6 and Kimi K3 with a Claude Code-oriented workflow is a signal that the coding harness and the inference model are becoming separable layers.

A practitioner setup combining GPT-5.6 and Kimi K3 with a Claude Code-oriented workflow is a signal that the coding harness and the inference model are becoming separable layers. The 5 source links point in the same direction: model choice is moving behind roles, gateways, subagents, and evaluation loops.

This is not evidence that every model is a drop-in replacement or that Anthropic or OpenAI endorses a particular cross-provider configuration. It is evidence that developers increasingly value the surrounding operating system: tool calls, repository context, approvals, planning, subagents, diffs, and verification.

Official product documentation points in the same architectural direction. Claude Code documents model routing through LLM gateways. Codex documents per-agent model and reasoning settings. Uber's uReview reports a production evaluation where one model generated review comments and another graded them.

## What is model mixing actually for?

The useful unit is not "best model." It is a role with a measurable success condition:

- fast model for repository exploration;
- deeper model for architecture or ambiguous debugging;
- independent model for critique or grading;
- specialized visual model for interface work;
- deterministic code for validation and state changes.

Model diversity only helps when outputs meet at an explicit contract. Without shared schemas, tests, and provenance, a multi-model setup becomes a collection of incompatible opinions.

## New Runtime Read

Our project already has the beginnings of this separation: deterministic ingestion, hosted structured analysis, Hermes for fast operations, local models for backlog, and OpenClaw for heavy research and owner-gated publication.

The next step is not adding models for novelty. It is measuring which role benefits from a different model and keeping the harness portable. The durable asset is the evaluated workflow; models should remain replaceable components inside it.