Repository navigation
Default to gpt-6-luna with low reasoning effort; release 0.8.0 - #69
Merged
Merged
Conversation
The "none" default broke the most common next step for a beginner: switching to a smarter model. gpt-6.1-sol and gpt-6-astra reject "none" with a bare 400, so `chat.model = "gpt-6.1-sol"` failed until you also changed reasoning_effort. Every current flagship model accepts "low", so changing the model alone now works. gpt-6-luna replaces gpt-5.6-terra as the default, including in generate_schema!, because it costs about a twentieth as much per token and answered a sample prompt about twice as fast. Even at "low", it beat terra at "none" (about 5s vs 7.5s). A non-"none" effort also requests a reasoning summary, so one is now requested by default. Set reasoning_effort = "none" for the old behavior on models that support it.
Gets `standardrb` back to a clean run so a new offense stands out instead of hiding among old ones. Every fix is behavior-neutral: the proxy= rescue still restores @proxy and re-raises, generate_schema! uses ||= for the same fallback, and the verbosity example drops unused assignments and fixes its indentation.
raghubetina
force-pushed
the
default-gpt-6-luna-low-effort
branch
from
October 19, 2026 15:54
ff61fbd to
79016b8
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
reasoning_effort: "none"broke the most common next step for a beginner, switching to a smarter model:gpt-6.1-solandgpt-6-astrareject"none"with a 400, sochat.model = "gpt-6.1-sol"failed until you also changed the effort."low"is accepted by all current flagship models (gpt-6-luna,gpt-6.1-sol,gpt-6-astra).gpt-6-lunareplacesgpt-5.6-terraas the default (and ingenerate_schema!): about 1/20 the per-token price, and roughly twice as fast on a sample prompt. Three runs each of "What's the best pizza in Chicago?":nonelowgpt-5.6-terragpt-6-lunaSo the new default (luna/low, ~5s) is faster than the old one (terra/none, ~7.5s).
Three commits: the change, a behavior-neutral
standardrbcleanup, then the 0.8.0 bump (mirroring #67/#68).Testing
bundle exec rakewith a liveOPENAI_API_KEY: 117 examples, 0 failures.examples/all.rbagainst the live API. One pre-existing mismatch unrelated to this change:07_edge_cases.rbexpectsBadRequestErrorfor an unknown model, but OpenAI now returnsNotFoundError(404).gpt-6.1-solwith no effort set (previously a 400).standardrbnow passes cleanly (was 10 offenses onmain).reekstill reports its existing design-smell warnings; those need refactors and are left for a separate change.