DRAFT: New post — Opus 4.8, Sonnet 5, and Fable 5 on GHA-bench - #2312
Open
Adam-S-Daniel wants to merge 7 commits into
Open
DRAFT: New post — Opus 4.8, Sonnet 5, and Fable 5 on GHA-bench#2312Adam-S-Daniel wants to merge 7 commits into
Adam-S-Daniel wants to merge 7 commits into
Conversation
Draft post (published: false) summarizing the opus-4.8 GHA-bench results: quality leader on tests + code (panel agrees, Spearman +0.83/+0.90) at a speed/cost premium (steepest at medium ~+65% vs 4.7, compressing to ~+15% at high; D/D- at xhigh/ultra), the new multi-agent "ultra" effort, and the trap caveat (~2x traps but ~99% benign per the 201-occurrence forensic). Notes the agy judge-harness and CC-version caveats. Links the updated sorting widget. 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 Deployed
Preview updates automatically on every push to this PR. |
Replace the link-only 'Try it yourself' section with the full interactive best-weighted-sorting table (Opus 4.8 + comparison models, one pooled calibration) plus the five weighting presets. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qx8jvUUZeg6DY1nxBT4pKf
Widget regenerated from the current 6-run combined report: 76 rows, geometric-mean duration/cost, powershell-tool pooled into pwsh, and sonnet 5 + fable 5 + split opus-4.6/sonnet-4.6 effort rows added. Duration cell now shows a slowest-run tooltip with timeout censoring (dagger marker + footnote). Prose: cost premiums re-derived (+54%/+67% at medium, ~+30% at high, ~+75-80% at xhigh), Spearman correlation range updated, and a new "how to read the numbers" methodology paragraph added.
Consolidates the two in-flight GHA-bench posts into a single draft (per review): renamed/retitled/redated, new intro covering all three campaigns (140 + 100 + 56 runs), new Sonnet 5 section (effort-tiered value story incl. the six 30-min timeouts at high) and Fable 5 section (near-top tests at medium, double-Opus pricing, clean sheet, weak free-choice-language quirk), presets called out in the widget intro, CC-version footnote extended. Widget payload unchanged (already covers all models). published: false stays.
Adam-S-Daniel
changed the title
Adam-S-Daniel
marked this pull request as ready for review
July 7, 2026 03:32
The detector bug the post described in present tense was fixed in GHA-bench #27 (PR #37, 2026-07-03); published trap numbers already exclude it. Recomputed the headline ratio with the corrected detector: 0.97 vs 0.44 firings per run at matched efforts (still ~2x), now quoted explicitly, with the artifact reframed as a finding of the hand review that has since been fixed. CC-version caveat retained.
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.
One draft post covering all three new GHA-bench models — Opus 4.8, Sonnet 5, and Fable 5 — featuring the updated interactive weighted-sorting table with presets. (
published: false— Adam flips it when ready.)Consolidates the previously separate widget-refresh PR #2311 (now closed) and the Opus-4.8-only draft into this single post, per review.
The post
_posts/2026-07-06-opus-4-8-sonnet-5-fable-5-on-gha-bench.md(new slug/title/date/excerpt to match the broader scope).The widget (unchanged from the previous revision of this branch)
76 rows regenerated from the current flagship 6-run combined report: geometric-mean duration/cost,
powershell-toolpooled intopwsh(4 language columns), Sonnet 5 / Fable 5 / split-effort 4.6 rows included, Max-Duration tooltip with≥censoring and † footnote, preset buttons.All prose figures verified against
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.🤖 Generated with Claude Code
https://claude.ai/code/session_01N1CskCMscspn3CfVjXuz2C