DRAFT: GHA-bench post — refresh sorting widget with Opus 4.8 data - #2311
Closed
Adam-S-Daniel wants to merge 3 commits into
Closed
DRAFT: GHA-bench post — refresh sorting widget with Opus 4.8 data#2311Adam-S-Daniel wants to merge 3 commits into
Adam-S-Daniel wants to merge 3 commits into
Conversation
Rebuild the best-weighted-sorting widget ROWS from the current canonical cross-run report (opus-4.8 + opus-4.7 + sonnet-4.6 + opus-4.6 + haiku-4.5), all on one pooled calibration so grades compare across models. Adds Opus 4.8 at medium/high/xhigh + the new "ultra" (multi-agent) effort. Notes the judge harness change (Gemini via Antigravity `agy`, ~0.3 stricter) and that Opus 4.8 sometimes picks JS/PowerShell for the free-choice language. Post stays published: true (update to a live post); review before merging. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qx8jvUUZeg6DY1nxBT4pKf
github-actions
Bot
added
not-decap-created
Contributor
|
🗑️ Preview environment cleaned up. |
Five one-click presets (Balanced, Max quality, Quality on a budget, Cheapest, Fastest) that set all four weights at once; manual slider edits clear the active preset. Theme-neutral pill buttons. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qx8jvUUZeg6DY1nxBT4pKf
Owner
Author
|
Closing per review: consolidated into a single draft post covering Opus 4.8, Sonnet 5, and Fable 5 with the updated presets widget — see PR #2312. The published 2026-05-12 post keeps its current widget; the new post carries the refreshed 76-row table. |
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
2 participants
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.
Draft — do not merge until reviewed. Updates the live "Introducing GHA-bench" post's interactive best-weighted-sorting widget.
What changed
ROWSfrom the current canonical cross-run report (results_2026-06-26_103905__2026-05-06_173435__2026-04-17_004319__2026-04-09_152435.md), so the table now includes Opus 4.8 at medium / high / xhigh + the new "ultra" (multi-agent orchestration) effort — alongside Opus 4.7 (1m/200k), Sonnet 4.6 (1m/200k), Opus 4.6, and Haiku 4.5. 60 rows, all on one pooled calibration so the letter grades compare across models.agyCLI (replaced the retired Gemini CLI, June 2026);agygrades ~0.3 pts stricter (r ≈ 0.90 with the old harness), so the newest rows are if anything slightly conservative.Notes for review
published: true(this updates a live post; settingpublished: falsewould unpublish it). The--draftstatus is what prevents auto-merge. If you'd actually prefer it unpublished while iterating, say so and I'll flip it.main(PR "Claude PR Assistant workflow" #28).🤖 Generated with Claude Code
Update 2026-07-06: synced with the geomean report overhaul (GHA-bench PR #38)
Commit
16c6044fre-syncs the widget with the current flagship 6-run combined report (results_2026-06-30_191904__2026-07-01_184135__2026-06-26_103905__2026-05-06_173435__2026-04-17_004319__2026-04-09_152435.md):powershell-toolis pooled intopwsh(4 language columns, GHA-bench Add visual regression approval gate and admin review dashboard #30); adds Sonnet 5 (low/med/high), Fable 5 (med/high), and split-effort Opus 4.6 / Sonnet 4.6 rows.≥= right-censored lower bound).🤖 Generated with Claude Code
https://claude.ai/code/session_01N1CskCMscspn3CfVjXuz2C