Repository navigation
E2E S9: use reasoning models (tiered), not the cheap chat tier - #29
Merged
Merged
Conversation
The multi-model comparison should exercise REASONING models — that's the AI mode's real value — tiered by cost per the run decision: * OpenAI → o3 (the only reasoning model; not gpt-5 chat) — $2/$8 * Claude → claude-sonnet-4-6 (mid-tier reasoning, NOT Opus $15/$75) — $3/$15 * Gemini → gemini-3-pro (reasoning) — $2/$12 * DeepSeek → deepseek-reasoner (frontier R1, cheap) — $0.28/$0.42 pickModel now: env override (E2E_<PROVIDER>_MODEL) → the tiered reasoning default if the catalog exposes it → any reasoning model the provider lists → first. Each model is swappable without code changes. Reasoning models emit more output (hidden CoT), so the per-run estimate rose from "a few cents" to ≈ €0.01 (deepseek) … €0.22 (sonnet); all four + pause/resume ≈ €0.5–0.75, hard-capped per run at max_eur. Doc cost table updated. Still gated on E2E_LLM + keys; compiles + skips cleanly without them. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Follow-up to #28. The multi-model comparison (S9) should exercise reasoning models — the AI mode's real value — tiered by cost:
o3(reasoning; not gpt-5)claude-sonnet-4-6(mid-tier reasoning, not Opus @ 15/75)gemini-3-pro(reasoning)deepseek-reasoner(frontier R1)pickModelresolves:E2E_<PROVIDER>_MODELoverride → the tiered reasoning default (if in the catalog) → any reasoning model → first. Reasoning models emit more output (hidden CoT), so the estimate rose to ≈ €0.01 (deepseek) … €0.22 (sonnet); all four + pause/resume ≈ €0.5–0.75, hard-capped per run atmax_eur. Compiles + skips cleanly without keys.🤖 Generated with Claude Code