feat(app): opus 4.8[1m] picker entry + per-model effort gating#1373
Open
jlixfeld wants to merge 2 commits into
Open
feat(app): opus 4.8[1m] picker entry + per-model effort gating#1373jlixfeld wants to merge 2 commits into
jlixfeld wants to merge 2 commits into
Conversation
Anthropic released Claude Fable 5 (`claude-fable-5`) as a GA model on 2026-06-09 — a Mythos-class, most-capable model for long-horizon agentic work (1M context, adaptive thinking always-on, no extended thinking). Expose it in both Claude model pickers: - codium Anthropic plugin: add `claude-fable-5` (full-ID list). - happy-app Claude fallback: add `claude-fable-5` to getClaudeModelModes. Fable 5 is a new family with no short alias (unlike opus/sonnet/haiku), so the picker sends the full model id; the Claude CLI passes it through to the API unchanged. The live model list still comes from the CLI's session metadata when present — this is the offline fallback. Add a getClaudeModelModes test asserting the fallback list and that the full id is used for Fable 5. Generated with [Claude Code](https://claude.ai/code) via [Happy](https://happy.engineering) Co-Authored-By: Claude <noreply@anthropic.com> Co-Authored-By: Happy <yesreply@happy.engineering>
Two related model-picker improvements for Claude:
1. Expose `claude-opus-4-8[1m]` (the 1M-context Opus 4.8 variant) as a
distinct picker entry in both the happy-app Claude fallback and the
codium Anthropic plugin. It has no short alias, so the full model id is
sent and passed through to the API unchanged.
2. Gate effort levels per model. Previously getEffortLevelsForModel ignored
the model and returned the full claude set for every model — which offers
levels the API rejects (effort errors entirely on Haiku 4.5; `xhigh` is
Opus 4.7+/Fable only; `max` is not on Opus 4.5). Per the Anthropic effort
docs:
- Fable 5 / Opus 4.8 (+ 1M) / Opus 4.7: low, medium, high, xhigh, max
- Sonnet 4.6 / Opus 4.6: low, medium, high, max
- Opus 4.5: low, medium, high
- Haiku 4.5: none
Unknown/custom (gateway) models keep the full set since support can't be
inferred. Matching works for both picker aliases (`opus`/`sonnet`) and
full model ids from gateway/SDK metadata.
Add tests covering the new model list and the per-model effort matrix.
Generated with [Claude Code](https://claude.ai/code)
via [Happy](https://happy.engineering)
Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Happy <yesreply@happy.engineering>
jlixfeld
added a commit
to jlixfeld/happy
that referenced
this pull request
Jun 9, 2026
jlixfeld
added a commit
to jlixfeld/happy
that referenced
this pull request
Jun 9, 2026
jlixfeld
added a commit
to jlixfeld/happy
that referenced
this pull request
Jun 9, 2026
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
1 participant
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.
What
Two Claude model-picker improvements:
1.
claude-opus-4-8[1m]picker entryAdds the 1M-context Opus 4.8 variant as a distinct, selectable entry in both the happy-app Claude fallback (
opus 4.8 (1M)) and the codium Anthropic plugin. No short alias exists, so the full model id is sent and passed through to the API unchanged.2. Per-model effort gating
getEffortLevelsForModelpreviously ignored the model and returned the full Claude effort set for every model — offering levels the API rejects. Per the Anthropic effort docs:This fixes a latent bug: the picker offered
xhigh/maxfor Haiku (effort errors entirely on Haiku) andxhighfor Sonnet 4.6 (unsupported). Unknown/custom gateway models keep the full set (support can't be inferred). Matching handles both picker aliases (opus/sonnet) and full model ids from gateway/SDK metadata.Verification
modelModeOptions.test.ts— 11 passed (new model-list + per-model effort matrix tests)happy-apptypecheck — passcodiumtypecheck — passSmoke-tested live in the running app via the
dev/session-composerpicker:opus 4.8 (1M)entry present:Fable 5 offers
xhigh(full effort set):Haiku 4.5 hides the effort control entirely (no effort param):
Sonnet 4.6 was confirmed to cycle low/medium/high/max with no
xhigh.The relabel-to-4.8 (#1345) and
xhigheffort level (#1269) work this builds on is already inmainvia #1363.🤖 Generated with Claude Code