🤖 feat: promote Claude Mythos 5.1 as the mythos model (DO NOT MERGE until GA)

[ThomasK33]

Summary

Prepares Xum for a rumored Claude Mythos 5.1 drop: promotes Mythos 5.1 (anthropic:claude-mythos-5-1) to the MYTHOS known model, so the mythos alias and model picker route to the new id. Mythos 5 stays usable as the custom model string anthropic:claude-mythos-5 and keeps its metadata entry and tokenizer approximation. This is the Mythos twin of the Fable 5.1 promotion (#3988) and is stacked on mike/fable-5-1-drop.

Important

Merge gate: hold until Anthropic officially announces Mythos 5.1, then verify the assumptions below against the announcement before merging. Nothing is announced yet — merging early would break mythos-alias sends (the API id would not exist yet).

Background

Fable 5 and Mythos 5 shipped together in ba36c81; the Fable 5.1 prep (#3988) covers only the Fable side. This PR is the forgotten Mythos counterpart, prepared ahead of any announcement so a reviewer can fix the numbers in five minutes when the real spec lands. Sibling PRs prep Opus 5.1 (opus-51-prep) and Sonnet 5.1 (sonnet-51-prep) on the same stack.

Assumptions (every one is a guess — verify on announcement day)

• That Mythos 5.1 exists at all: inferred from Fable 5 and Mythos 5 shipping as twins; Anthropic may not ship a 5.1 for the restricted-access variant.
• API id: claude-mythos-5-1 — invented by analogy to claude-fable-5-1 (🤖 feat: promote Claude Fable 5.1 as the fable model #3988) and the dateless dash convention of the 5 generation.
• Pricing: $10/M input, $50/M output, cache write 1.25× input, cache read 0.1× input — copied from Mythos 5 (which was itself identical to Fable 5, and unchanged for Fable 5.1).
• Context/output envelope: native 1M context, 128K max output — copied from Mythos 5.
• Thinking/effort: full 5-level ladder (low…max) with native xhigh, adaptive thinking only (disabled thinking rejected, "off" clamps to "low") — copied from Mythos 5 via the existing claude-(?:fable|mythos)- wildcard matchers.
• Tokenizer: Opus 4.5 approximation (anthropic/claude-opus-4.5) — copied from Mythos 5 / Fable 5.1; the real Mythos-class tokenizer is unpublished, so counts can run ~1.0–1.3× low.
• Availability: still restricted-access (no self-serve) — copied from Mythos 5; this drives the deliberate omissions below.

Implementation

Mirrors #3988 surface-for-surface, adapted to Mythos:

• knownModels.ts: MYTHOS → claude-mythos-5-1 (keeps the mythos alias; stays unwarmed since most users cannot access it and its tokenizer override is already warmed via FABLE). LEGACY_TOKENIZER_MODEL_OVERRIDES retains anthropic:claude-mythos-5 → Opus 4.5 approximation, exactly like the retired claude-fable-5 / claude-opus-4-8 ids.
• models-extra.ts: new claude-mythos-5-1 pricing/limits entry; the Mythos 5 entry is retained.
• models.ts: claude-mythos-5-1 added to the native-1M patterns, anchored like Mike's (^claude-mythos-5-1 + optional date suffix) because the existing ^claude-mythos-5 pattern tolerates date suffixes but not a -1 minor.
• No wire-format changes needed for thinking/effort/web-fetch: the Mythos-class wildcard matchers (anthropicSupportsNativeXhigh, anthropicRejectsDisabledThinking) and native web-fetch id parsing already match the new id; tests now pin that.
• Docs model table and built-in skill content regenerated (make fmt + generate-builtin-skills.sh), not hand-edited.

Deliberate omissions vs #3988

• No defaultModelFallbacksSeededMythos51 migration flag and no config.ts/appConfigOnDisk.ts changes: DEFAULT_MODEL_FALLBACKS has no Mythos source key — the refusal-fallback chain exists because Fable's safeguards refuse requests Opus can serve, and Mythos (safeguards lifted) never had a default chain to re-seed. Adding one would invent new default behavior beyond the promotion; adding only the flag would be dead code. Because this PR does not touch the migrations schema, it trivially coexists with defaultModelFallbacksSeededFable51 and the Opus/Sonnet flags from the sibling preps, and downgrade behavior is unchanged.
• No WorkflowRunner fixture change: no fixture pins the mythos alias (audited; the Fable one Mike updated was the only alias-pinning fixture).

Every fable-5-1 touchpoint on Mike's branch was grepped and either mirrored or listed above as deliberately omitted; remaining claude-mythos-5 references in production code are the intentional legacy entries (metadata, tokenizer override, 1M pattern for the old id) plus comments.

Validation

• Behavioral coverage added for the new id: native-1M classification (direct + gateway-prefixed), display formatting (Mythos 5.1), 5-level thinking policy with off-clamp, provider options (adaptive + summarized display, no disabled thinking), and native web-fetch support.
• make static-check green; targeted suites (models, modelDisplay, policy, providerOptions, tools: 360 tests) and config.test.ts (174 tests, confirming the untouched seed machinery) pass.

Risks

Low while gated: additive registry/metadata changes only. The main user-visible effect after merge is that the mythos alias and picker route to Mythos 5.1; until the model exists at the API, such sends would fail — which is why the merge gate matters. Existing Mythos 5 selections keep working as custom model strings with unchanged token accounting (the legacy tokenizer override resolves to the same Opus 4.5 approximation). If the announced spec differs, the fix is confined to the claude-mythos-5-1 entry in models-extra.ts and the MYTHOS definition in knownModels.ts.

---

Generated with xum • Model: anthropic:claude-fable-5 • Thinking: xhigh

[ThomasK33]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Nice work!

Reviewed commit: 35f9ba849e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Security review completed. No security issues were found in this pull request.

Reviewed commit: 35f9ba849e

View security finding report

Only the user who started this review can view the report in Codex.

ℹ️ About Codex security reviews in GitHub

This is an experimental Codex feature. Security reviews are triggered when:

• You comment "@codex security review"
• A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review

Once complete, Codex will leave suggestions, or a comment if no findings are found.