Coding router comparison

Weave Router vs Not Diamond

Both products make model decisions during coding sessions. Compare Weave Router and Not Diamond on session policy, cache behavior, deployment and total cost. Weave also offers separate output and AI-spend analysis. No common benchmark establishes a performance advantage between them.

Explore Weave RouterSee pricing

Sources reviewed

The short version

Compare deployment, session policies and output analysis.

Weave Router

Inspectable routing and separate analytics

  • Inspect or self-host the public routing implementation
  • Evaluate cache-aware switching and escalation across sessions
  • Add separately offered output and AI-cost analysis
54786
What is Weave Router?

Weave Router selects coding models using complexity, cache economics and escalation. Use the hosted installer or operate the open-source stack. Its separate intelligence products connect output estimates, quality, AI attribution and cost.

Not Diamond

Metadata-based coding recommendations

  • Recommend models and reasoning effort per step
  • Use cache state, feedback and prior turns
  • Send model requests through your provider or gateway
23
What is Not Diamond?

Not Diamond Code routes long-horizon coding sessions through a local proxy and metadata-based optimization service. It recommends a model and reasoning effort per step; requests use your configured provider or gateway.

Side by side

Weave Router and Not Diamond, compared

What mattersWeave RouterNot Diamond
Coding-agent support523Installer targets Claude Code, Codex, OpenCode and Pi. Cursor is documented in early beta.Not Diamond Code is built for long-horizon coding agents. Its pricing FAQ explicitly confirms Claude Code support.
Session-level decisions43Documents complexity-based model selection, cache-aware switching, escalation and supported subscription routing.Recommends a model and reasoning effort using payload semantics, cache state, feedback, session outcomes and earlier turns.
Setup and deployment52Hosted installer or an inspectable self-hosted router stack. Self-hosting requires operating the service and its database.Local proxy computes optimization metadata; requests use your provider or gateway. Access is currently gated by an early-access waitlist.
Data path52Review the hosted or self-hosted credential and request path in the public source before rollout.Code docs say raw prompts, code, inputs and outputs are not sent to Not Diamond; derived metadata is used for recommendations.
Pricing meter463Hosted Router plans, provider usage and self-hosting have different costs. Review Router terms separately from the Weave platform subscription.
  • Lists $0.05 per million tokens routed on pay-as-you-go.
  • That routing fee is separate from model inference.
  • Enterprise terms and volume discounts require a quote.
Routing evidence523Public routing implementation and benchmark harness can be inspected. The published Router evaluation is not a head-to-head Not Diamond test.Documents the session inputs and optimization objective. Validate vendor savings claims on the configuration and tasks you would run.
From spend to output78963Separate platform access adds Silk 1 expert-effort estimates, team calibration, AI attribution and cost analysis. Not included in Router plans.Savings and usage dashboard; organization-wide analytics in Enterprise. Inspect metrics rather than treating usage as accepted output.
Sources and comparison method

The key difference

Compare the session policy and the outcome you need to explain

Weave Router

Weave’s approach

Inspect Weave’s routing implementation, then evaluate its separate intelligence products for output estimates, AI attribution and spend. Confirm coverage and billing for each product.

2357896

Not Diamond

Not Diamond’s approach

Not Diamond Code uses expected downstream work, cache state and usage telemetry to optimize session quality and cost. Its local proxy sends derived metadata. It supports coding agents, not just application APIs.

2357896

Test both policies on the same work. Evaluate Weave’s separate output and cost methods if you also need engineering ROI analysis.

Why Weave Router

What makes Weave worth evaluating

Inspect the routing implementation

Review the public source, supported models and configuration. Compare a pinned model with automatic routing on representative repository tasks.

5

Account for the whole coding session

Weave’s documented policy includes cache-aware switching and escalation. Evaluate completed tasks, retries and cache charges together. These policy features do not prove a lower bill than Not Diamond.

43

Connect cost with explained output

Silk 1 estimates expert effort and supports team calibration. Token Intelligence adds AI attribution and cost sources. Check the published method’s units, coverage and assumptions. These platform capabilities are offered separately from Router.

7896

In practice

Evaluate the workflow you will actually run

Keep the session context

Weave Router

Inspect enabled models, cache decisions, escalations and supported subscription behavior across an entire coding task.

Not Diamond

Inspect recommendations across prior turns, cache state, reasoning effort and compaction events.

453

Reconcile the bill

Weave Router

Separate the hosted plan or self-hosting cost from provider usage and the platform products you add.

Not Diamond

Separate the published routing fee from model inference and any gateway or enterprise charges.

463

Explain the result

Weave Router

Inspect the accepted code and, with the platform, the output estimate, attribution and cost coverage.

Not Diamond

Inspect accepted code alongside the savings and usage dashboard. Ask which baseline and cost categories underpin reported savings.

7893

Choose for your workflow

When Not Diamond may fit your team

Evaluate Not Diamond for metadata-based recommendations alongside your provider or gateway. Confirm early access, supported harness versions and commercial terms. Both products support coding agents; compare routing policy, operating model and the outcome evidence you need.

235

Try it on your own work

Evaluate a switch before changing the default

  1. Fix the comparison conditions

    Use the same repository revision, task instructions, agent version and acceptance checks. Record the available models and credential path for each configuration.

  2. Measure completed work

    Record pass rate, elapsed time and total cost. Include failed tasks, retries, cache rebuilds, routing fees and classifier overhead where applicable. Do not infer savings from token price alone.

  3. Roll out with a fallback

    Check tool calls, streaming, context limits and data handling in a small pilot. Retain a known working configuration and broaden adoption after reviewing the results.

A few more details

Frequently asked questions

All comparisons

Evaluate routing against the work it produces

Inspect Weave Router’s policy and setup, then test it on representative coding tasks.

Explore Weave RouterSee pricing

Behind the comparison

Sources and comparison method

Published by Weave. First-party sources reviewed October 16, 2026. This is a comparison of documented capabilities and workflow fit. No common benchmark of Weave Router and Not Diamond was run. Vendor savings claims do not establish a measured advantage between the products.

  1. Not Diamond: model routing for coding agents
  2. Not Diamond Code: local proxy, metadata and early access
  3. Not Diamond: routing fee, session inputs and enterprise terms
  4. Weave Router: complexity, cache economics and escalation
  5. Weave Router: public implementation, installer and self-hosting
  6. Weave pricing: platform and separate Router plans
  7. Code Intelligence: Silk 1 output estimates and calibration
  8. Token Intelligence: AI attribution and cost sources
  9. Weave AI ROI: output units, cost and coverage limits

Inspect the evidence