rightmodeler — prove which models you can safely downgrade

Evidence-backed model downgrades<br>Prove which models you can safely downgrade.<br>rightmodeler replays your real agent traces through cheaper models, judges each output against what you already shipped, and shows exactly where you can cut cost — with evidence and confidence on every call.<br>0%cost reduction · PR summary · medium confidence — illustrative example

$uv run python -m pipeline ingestView on GitHub

Illustrative — not measured results<br>cheaper-models · per-step approval5 steps · quality floor 0.90<br>✓Approvedpr_summary<br>gpt-4.1 → gpt-4o-mini<br>save 72%quality 0.94reference+judge

⚙Pending — cascade risktool_agentCASCADE<br>claude-opus-4 → llama-3.3-70b<br>save 41%quality 0.88trajectory

✓Approvedjson_extraction<br>gpt-4o → gpt-4o-mini<br>save 68%quality 1.00deterministic

✓Approvedsql_generation<br>gpt-4o → deepseek-chat<br>save 55%quality 0.91reference

✕Abstainedauth_code_editHIGH-RISK · abstain<br>gpt-4.1 → —<br>save —quality —none

statusStepFamilyCurrent → CandidateSaveQualityEvidenceFlag<br>✓Approved1pr_summarygpt-4.1 → gpt-4o-mini72%0.94reference+judge—<br>⚙Pending — cascade risk2tool_agentclaude-opus-4 → llama-3.3-70b41%0.88trajectoryCASCADE<br>✓Approved3json_extractiongpt-4o → gpt-4o-mini68%1.00deterministic—<br>✓Approved4sql_generationgpt-4o → deepseek-chat55%0.91reference—<br>✕Abstained5auth_code_editgpt-4.1 → ———noneHIGH-RISK · abstain

Reads the traces you already have.<br>Autodetects 8 trace formats into one per-step schema.

Reads traces from Claude Code, Codex, LangSmith / LangGraph, OpenAI SDK, Langfuse, Braintrust, Phoenix (OpenInference), OTel GenAI.<br>Claude CodeCodexLangSmith / LangGraphOpenAI SDKLangfuseBraintrustPhoenix (OpenInference)OTel GenAIClaude CodeCodexLangSmith / LangGraphOpenAI SDKLangfuseBraintrustPhoenix (OpenInference)OTel GenAI

Claude CodeCodexLangSmith / LangGraphOpenAI SDKLangfuseBraintrustPhoenix (OpenInference)OTel GenAIClaude CodeCodexLangSmith / LangGraphOpenAI SDKLangfuseBraintrustPhoenix (OpenInference)OTel GenAI

Run it on your own traces.

$uv run python -m pipeline ingest# It’s a report, not a runtime gateway — prove the savings on your own data first.

View on GitHub

Every figure shown on this page is an illustrative example — not measured savings.<br>cheaper-models · © 2026 rightmodeler · an ELM-OS project