aaronjmars · Oct 7, 2026 · Oct 7, 2026
diff --git a/.github/workflows/aeon.yml b/.github/workflows/aeon.yml
@@ -34,9 +34,10 @@ on:
           - '(config default)'
           - claude-sonnet-5-5
           - claude-opus-5-5
-          - claude-haiku-4-5-20251001
+          - claude-haiku-5-5
           - claude-sonnet-5
           - claude-opus-4-8
+          - claude-haiku-4-5-20251001
           - grok-4.7
           - grok-4.6
           - grok-4.5
@@ -1750,7 +1751,7 @@ jobs:
 
           JSON format: {\"score\": N, \"assessment\": \"one line\", \"flags\": []}"
 
-          SCORE_MODEL="${ANTHROPIC_DEFAULT_HAIKU_MODEL:-claude-haiku-4-5-20251001}"
+          SCORE_MODEL="${ANTHROPIC_DEFAULT_HAIKU_MODEL:-claude-haiku-5-5}"
           # Score through the SAME harness the skill ran on, ALWAYS via the
           # harness-adapter — no native `claude -p` / `run-grok.sh` path remains.
           # A repo authed for only one harness has no Claude creds, so a direct

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -176,6 +176,13 @@ from or pin to; the template keeps serving the latest `main` to new forks.
 
 ### Changed
 
+- **Haiku tier moves to Claude Haiku 5.5.** `claude-haiku-5-5` replaces `claude-haiku-4-5-20251001`
+  in the dashboard picker, the workflow_dispatch choices, the quality scorer's default, the
+  json-render notifier and the docs; `claude-haiku-4-5-20251001` stays a valid dispatch choice and
+  shows as "Haiku 4.5 (configured)". The OpenRouter haiku slot defaults to
+  `anthropic/claude-haiku-5.5`; Surplus does not serve it yet, so its haiku tier stays on
+  `claude-haiku-4.5`. Fleet scorecard prices Haiku 5.5 at $0.10 in / $0.50 out per 1M tokens.
+
 - **Operator-console plugin fills its Claude directory listing.** `plugin/.claude-plugin/plugin.json`
   now sets `displayName`, author email, `icon` and the docs, support, privacy and terms links that
   Anthropic's plugin directory reads for the listing, and a shorter `description` that fits the

diff --git a/aeon.yml b/aeon.yml
@@ -2,7 +2,7 @@
 #
 # Enable skills by setting enabled: true. Multiple skills at the same time run in parallel.
 # Set var: "value" to pass a default parameter to the skill.
-# Set model: "claude-haiku-4-5-20251001" per skill to override the default model (cost optimization).
+# Set model: "claude-haiku-5-5" per skill to override the default model (cost optimization).
 # Set harness: "<name>" per skill (claude, grok, codex, pi, vibe, kimi, fx, cursor,
 # hermes) to run it on a different coding-agent CLI than the default Claude Code
 # harness (see the top-level `harness:` key below).
@@ -229,7 +229,7 @@ chains:
   #     - { skill: digest, consume: [token-movers, github-trending] }
 
 # Default model for all skills. Override per-run via workflow_dispatch.
-# Options: claude-sonnet-5-5, claude-opus-5-5, claude-haiku-4-5-20251001
+# Options: claude-sonnet-5-5, claude-opus-5-5, claude-haiku-5-5
 # (older claude-sonnet-5 / claude-opus-4-8 still accepted)
 # (Grok harness model: grok-4.7, the flagship reasoning model and grok's current
 #  default, run with --no-subagents in CI; grok-4.6 is also offered, and an existing

diff --git a/apps/dashboard/lib/config.test.ts b/apps/dashboard/lib/config.test.ts
@@ -190,9 +190,9 @@ describe("updateModelInConfig", () => {
   });
 
   it("replaces an existing model", () => {
-    const updated = updateModelInConfig(FULL_YAML, "claude-haiku-4-5-20251001");
+    const updated = updateModelInConfig(FULL_YAML, "claude-haiku-5-5");
     const config = parseConfig(updated);
-    assert.equal(config.model, "claude-haiku-4-5-20251001");
+    assert.equal(config.model, "claude-haiku-5-5");
   });
 });
 

diff --git a/apps/dashboard/lib/constants.test.ts b/apps/dashboard/lib/constants.test.ts
@@ -73,7 +73,7 @@ describe("keyProvidedByHarness", () => {
 
 describe("MODELS", () => {
   it("offers exactly the current three Claude models", () => {
-    assert.deepEqual(MODELS.map(m => m.id), ["claude-sonnet-5-5", "claude-opus-5-5", "claude-haiku-4-5-20251001"]);
+    assert.deepEqual(MODELS.map(m => m.id), ["claude-sonnet-5-5", "claude-opus-5-5", "claude-haiku-5-5"]);
   });
 });
 

diff --git a/apps/dashboard/lib/constants.ts b/apps/dashboard/lib/constants.ts
@@ -6,11 +6,12 @@ import type { Harness } from './types'
 // harness-switch snap uses (modelsForHarness(...)[0] in app/page.tsx). Keep it in
 // sync with the config default in lib/config.ts and aeon.yml `model:`.
 // Just the current three. A repo or per-skill pin that still names an older id
-// (claude-sonnet-5, claude-opus-4-8) keeps a visible entry via pickerOptions.
+// (claude-sonnet-5, claude-opus-4-8, claude-haiku-4-5-20251001) keeps a visible
+// entry via pickerOptions.
 export const MODELS = [
   { id: 'claude-sonnet-5-5', label: 'Sonnet 5.5' },
   { id: 'claude-opus-5-5', label: 'Opus 5.5' },
-  { id: 'claude-haiku-4-5-20251001', label: 'Haiku 4.5' },
+  { id: 'claude-haiku-5-5', label: 'Haiku 5.5' },
 ]
 
 // Models offered when the Grok (`grok`) harness is selected: the ids the
@@ -137,6 +138,7 @@ export const HARNESSES = [
 const RETIRED_MODEL_LABELS: Record<string, string> = {
   'claude-sonnet-5': 'Sonnet 5',
   'claude-opus-4-8': 'Opus 4.8',
+  'claude-haiku-4-5-20251001': 'Haiku 4.5',
   'openai/gpt-5.1-codex-mini': 'GPT-5.1 Codex Mini',
   'openai/gpt-5-mini': 'GPT-5 Mini',
   'openai/gpt-5.3-codex': 'GPT-5.3 Codex',

diff --git a/apps/dashboard/lib/dispatch.test.ts b/apps/dashboard/lib/dispatch.test.ts
@@ -31,7 +31,7 @@ describe("sanitizeModel", () => {
   it("keeps dotted versions and colon variant suffixes", () => {
     assert.equal(sanitizeModel("grok-4.5"), "grok-4.5");
     assert.equal(sanitizeModel("vendor/model:free"), "vendor/model:free");
-    assert.equal(sanitizeModel("claude-haiku-4-5-20251001"), "claude-haiku-4-5-20251001");
+    assert.equal(sanitizeModel("claude-haiku-5-5"), "claude-haiku-5-5");
   });
 
   it("strips characters outside the id charset", () => {

diff --git a/docs/CONFIGURATION.md b/docs/CONFIGURATION.md
@@ -297,11 +297,11 @@ The default model for all skills is set in `aeon.yml` (or from the dashboard hea
 model: claude-sonnet-5-5
 ```
 
-Options: `claude-sonnet-5-5` (default), `claude-opus-5-5`, `claude-haiku-4-5-20251001` (the older `claude-sonnet-5` and `claude-opus-4-8` are still accepted). Per-run overrides are available via workflow dispatch, and individual skills can override to optimize cost:
+Options: `claude-sonnet-5-5` (default), `claude-opus-5-5`, `claude-haiku-5-5` (the older `claude-sonnet-5`, `claude-opus-4-8` and `claude-haiku-4-5-20251001` are still accepted). Per-run overrides are available via workflow dispatch, and individual skills can override to optimize cost:
 
 ```yaml
 skills:
-  token-movers: { enabled: true, schedule: "30 12 * * *", model: "claude-haiku-4-5-20251001" }
+  token-movers: { enabled: true, schedule: "30 12 * * *", model: "claude-haiku-5-5" }
 ```
 
 > Model ids are for the **claude** harness; each other [harness](harnesses.md) carries its own list, which the dashboard picker swaps in when you select it.

diff --git a/scripts/fleet-scorecard.mjs b/scripts/fleet-scorecard.mjs
@@ -118,13 +118,14 @@ const CLAUDE_PRICES = {
   'claude-sonnet-4-6': { in: 3, out: 15, cr: 0.30 },
   'claude-sonnet-4-5': { in: 3, out: 15, cr: 0.30 },
   'claude-sonnet-4': { in: 3, out: 15, cr: 0.30 },
+  'claude-haiku-5-5': { in: 0.10, out: 0.50, cr: 0.01 },
   'claude-haiku-4-5': { in: 1, out: 5, cr: 0.10 },
   'claude-3-5-haiku': { in: 0.80, out: 4, cr: 0.08 },
 };
 // A Claude id with a version this table does not know yet (a new release) is
 // priced at its family's current flagship row, so it still shows up in the
 // cost columns instead of vanishing.
-const CLAUDE_FAMILY_FALLBACK = { opus: 'claude-opus-5-5', sonnet: 'claude-sonnet-5-5', haiku: 'claude-haiku-4-5' };
+const CLAUDE_FAMILY_FALLBACK = { opus: 'claude-opus-5-5', sonnet: 'claude-sonnet-5-5', haiku: 'claude-haiku-5-5' };
 const normalizeModel = (model) => String(model || '').trim().toLowerCase()
   .replace(/^anthropic\//, '').replace(/-\d{8}$/, '').replace(/\./g, '-');
 // Rates in $ per token for a model, or null when it is not a Claude model. Rows

diff --git a/scripts/llm-gateway.sh b/scripts/llm-gateway.sh
@@ -262,7 +262,7 @@ case "${GATEWAY:-direct}" in
     # Map EVERY model slot Claude Code uses to OpenRouter slugs (opus/sonnet/haiku).
     export ANTHROPIC_DEFAULT_OPUS_MODEL="${OPENROUTER_MODEL:-anthropic/claude-opus-5.5}"
     export ANTHROPIC_DEFAULT_SONNET_MODEL="${OPENROUTER_MODEL_SONNET:-anthropic/claude-sonnet-5.5}"
-    export ANTHROPIC_DEFAULT_HAIKU_MODEL="${OPENROUTER_MODEL_HAIKU:-anthropic/claude-haiku-4.5}"
+    export ANTHROPIC_DEFAULT_HAIKU_MODEL="${OPENROUTER_MODEL_HAIKU:-anthropic/claude-haiku-5.5}"
     # Tiered mapping, same as the glm arm: the run's resolved model id picks the
     # slot, so sonnet-tier skills (and the scorer) stay on sonnet instead of every
     # run being billed as Opus.
@@ -360,8 +360,12 @@ X-Title: ${OPENROUTER_APP_TITLE:-Aeon}"
     # <digit>-<digit> to <digit>.<digit> (claude-opus-5-5 -> claude-opus-5.5).
     # SURPLUS_MODEL overrides; opus-5.5 is the fallback when $MODEL is unset.
     # Surplus served claude-opus-5.5 and claude-sonnet-5.5 on 2026-10-01
-    # (/api/inference/v1/models).
+    # (/api/inference/v1/models). It did not serve claude-haiku-5.5 yet on
+    # 2026-10-07, so the haiku tier stays on claude-haiku-4.5 there.
     surplus_model="${SURPLUS_MODEL:-$(printf '%s' "${MODEL:-claude-opus-5-5}" | sed -E 's/-[0-9]{8}$//; s/([0-9])-([0-9])/\1.\2/g')}"
+    if [ -z "${SURPLUS_MODEL:-}" ] && [ "$surplus_model" = "claude-haiku-5.5" ]; then
+      surplus_model="claude-haiku-4.5"
+    fi
     start_ccr_sidecar surplus \
       "https://www.surplusintelligence.ai/api/inference/v1/chat/completions" \
       "$SURPLUS_API_KEY" "$surplus_model"

diff --git a/scripts/notify-jsonrender.sh b/scripts/notify-jsonrender.sh
@@ -60,7 +60,7 @@ CRITICAL RULES:
 
 # Use claude CLI — works with both API key and OAuth token
 SPEC=$(echo "$CONTENT" | claude -p "Convert this skill output into a json-render spec. Skill: ${SKILL}" \
-  --model claude-haiku-4-5-20251001 \
+  --model claude-haiku-5-5 \
   --system-prompt "$SYSTEM" \
   --max-turns 1 \
   --output-format text 2>/dev/null)

diff --git a/scripts/tests/test_fleet_scorecard.mjs b/scripts/tests/test_fleet_scorecard.mjs
@@ -56,6 +56,7 @@ test('prices Claude rows by model version and leaves non-Claude rows unpriced',
     '2026-10-01,a,claude-sonnet-5,1000000,1000000,0,0', // 2 + 10 = 12
     '2026-10-01,a,claude-sonnet-4-6,1000000,1000000,0,0', // 3 + 15 = 18
     '2026-10-01,a,claude-haiku-4-5-20251001,1000000,1000000,1000000,1000000', // 1 + 5 + read 0.1 + write 1.25 = 7.35
+    '2026-10-01,a,claude-haiku-5-5,10000000,10000000,10000000,10000000', // 10M each at a tenth of 4.5's rates = 7.35
     '2026-10-01,b,openai/gpt-5.1-codex-mini,1000000,1000000,0,0', // unpriced
     '2026-10-01,b,codex-default,1000000,1000000,0,0', // unpriced
     '2026-10-01,b,grok-4.5,1000000,1000000,0,0', // unpriced
@@ -74,8 +75,8 @@ test('prices Claude rows by model version and leaves non-Claude rows unpriced',
   try {
     await import(new URL(`../fleet-scorecard.mjs?test=pricing-${Date.now()}`, import.meta.url))
     const metrics = JSON.parse(readFileSync('/tmp/fleet-scorecard/metrics.json', 'utf8'))
-    assert.equal(metrics.generations, 8)
-    assert.equal(metrics.est_cost_usd, 24 + 30 + 12 + 18 + 7.35)
+    assert.equal(metrics.generations, 9)
+    assert.equal(metrics.est_cost_usd, Math.round((24 + 30 + 12 + 18 + 7.35 + 7.35) * 100) / 100)
     assert.equal(metrics.unpriced_generations, 3)
     assert.equal(metrics.unpriced_tokens, 6_000_000)
     const body = readFileSync('/tmp/fleet-scorecard/scorecard-body.md', 'utf8')

diff --git a/scripts/tests/test_llm_gateway.sh b/scripts/tests/test_llm_gateway.sh
@@ -99,6 +99,9 @@ sidecar_model() {  # $1 = gateway, $2 = run model id ("" = unset); prints model=
 [ "$(sidecar_model surplus claude-haiku-4-5-20251001)" = "claude-haiku-4.5" ] \
   && pass "surplus: date suffix stripped, then dot-form" \
   || bad "surplus: date suffix stripped, then dot-form (got $(sidecar_model surplus claude-haiku-4-5-20251001))"
+[ "$(sidecar_model surplus claude-haiku-5-5)" = "claude-haiku-4.5" ] \
+  && pass "surplus: haiku-5-5 (not carried yet) stays on claude-haiku-4.5" \
+  || bad "surplus: haiku-5-5 (not carried yet) stays on claude-haiku-4.5 (got $(sidecar_model surplus claude-haiku-5-5))"
 [ "$(sidecar_model surplus "")" = "claude-opus-5.5" ] \
   && pass "surplus: unset MODEL falls back to opus-5.5" \
   || bad "surplus: unset MODEL falls back to opus-5.5 (got $(sidecar_model surplus ""))"
@@ -127,9 +130,12 @@ or_model() {  # $1 = run model id; prints the MODEL the arm resolves
 [ "$(or_model claude-opus-5-5)" = "anthropic/claude-opus-5.5" ] \
   && pass "openrouter: opus-pinned run gets the opus slug" \
   || bad "openrouter: opus-pinned run gets the opus slug (got $(or_model claude-opus-5-5))"
-[ "$(or_model claude-haiku-4-5-20251001)" = "anthropic/claude-haiku-4.5" ] \
+[ "$(or_model claude-haiku-5-5)" = "anthropic/claude-haiku-5.5" ] \
   && pass "openrouter: haiku-tier run gets the haiku slug" \
-  || bad "openrouter: haiku-tier run gets the haiku slug"
+  || bad "openrouter: haiku-tier run gets the haiku slug (got $(or_model claude-haiku-5-5))"
+[ "$(or_model claude-haiku-4-5-20251001)" = "anthropic/claude-haiku-5.5" ] \
+  && pass "openrouter: an older haiku id still lands on the haiku slot" \
+  || bad "openrouter: an older haiku id still lands on the haiku slot (got $(or_model claude-haiku-4-5-20251001))"
 ( export GATEWAY=openrouter OPENROUTER_API_KEY=test-key MODEL=claude-sonnet-5 \
     OPENROUTER_MODEL=x/opus OPENROUTER_MODEL_SONNET=x/sonnet
   # shellcheck disable=SC1090

diff --git a/skills/create-skill/SKILL.md b/skills/create-skill/SKILL.md
@@ -70,7 +70,7 @@ Today is ${today}. Your task is to generate a complete, production-ready skill f
    - **Variable behavior** — what `${var}` controls; what happens when empty (sane default OR clean abort with notify).
    - **Steps** — 4-8 numbered, following the standard pattern: read context → fetch/search → process/analyze → write output → log → notify.
    - **Schedule suggestion** — choose a cron slot. Read existing schedules in `aeon.yml`; avoid co-scheduling at the same minute as heavy skills (article, repo-scanner, deep-research, telegram-digest) unless the new skill is lightweight (<30s expected). Prefer a `:30` minute offset if the natural hour is already crowded.
-   - **Model** - default `claude-sonnet-5-5`. Pick `claude-haiku-4-5-20251001` if the skill is high-frequency aggregation/digestion (cost optimization), or `claude-opus-5-5` if it needs the strongest reasoning. Document the choice in the PR body.
+   - **Model** - default `claude-sonnet-5-5`. Pick `claude-haiku-5-5` if the skill is high-frequency aggregation/digestion (cost optimization), or `claude-opus-5-5` if it needs the strongest reasoning. Document the choice in the PR body.
    - **Category** - the pack the skill joins. Pick exactly one of `core` `evolution` `basics` `dev` `crypto` `productivity` (the set `scripts/check-skill-categories.sh` enforces; anything else, or a missing category, fails CI). For a new user-facing skill that is usually `basics`, `dev`, `crypto`, or `productivity`. See `docs/skill-packs.md`.
 
 6. **Write the SKILL.md draft** at `skills/{skill-name}/SKILL.md` with this exact structure:
@@ -145,7 +145,7 @@ Today is ${today}. Your task is to generate a complete, production-ready skill f
 
 9. **Register in `aeon.yml`.** Insert the new skill in the appropriate time-slot section:
    - Format: `  {skill-name}: { enabled: false, schedule: "{suggested_cron}" }`
-   - Add `model: "claude-haiku-4-5-20251001"` (or `"claude-opus-5-5"`) if chosen in step 5.
+   - Add `model: "claude-haiku-5-5"` (or `"claude-opus-5-5"`) if chosen in step 5.
    - Add `var: ""` if the skill takes a default var.
    - Add a brief trailing comment if the name doesn't make purpose obvious.
    - Place near related skills (crypto with crypto, content with content, etc.).

diff --git a/skills/skill-repair/SKILL.md b/skills/skill-repair/SKILL.md
@@ -122,7 +122,7 @@ Categories follow `CLAUDE.md`. Pick the **most specific** category that fits the
 |---|---|
 | **`api-change`** | WebFetch the live API spec / status page / release notes. Update endpoints, payload shape, headers, error codes in the skill. Cite the spec URL in the PR body. Never guess — if WebFetch fails, drop to `REPAIR_DIAGNOSED_NO_FIX`. |
 | **`rate-limit`** | Add backoff (`sleep`), reduce request count, or add a fallback endpoint. Never raise the limit from the skill side. If the skill's `schedule` is too aggressive, propose a less-frequent cron in the PR body but **don't edit `aeon.yml`** unless the issue file already authorizes it. |
-| **`timeout`** | Split work into stages, add early-return on partial success, downgrade `model:` to `claude-haiku-4-5-20251001` for the skill that doesn't need Sonnet or Opus. |
+| **`timeout`** | Split work into stages, add early-return on partial success, downgrade `model:` to `claude-haiku-5-5` for the skill that doesn't need Sonnet or Opus. |
 | **`sandbox-limitation`** | Usually the "sandbox blocks the network" myth — there is **no** network sandbox. The real cause is a bare `$SECRET` on the command line (refused by the Bash permission layer) or a non-allowlisted command. Fix: route auth-required calls through `./secretcurl` with a `{ENV_NAME}` placeholder, or `gh api` for GitHub (auth handled internally). **Irreversible side-effects** (email / spend / on-chain / deploy) run **in-run** via `./secretcurl` as the skill's final, fail-closed action — never for reads. Add/refresh a "Network note" section. (There are **no** `scripts/prefetch-*.sh` or `scripts/postprocess-*.sh` scripts — both patterns were retired; auth'd reads and irreversible sends alike happen in-run.) |
 | **`prompt-bug`** | Minimum-edit specificity insertion. Don't rewrite — add the missing constraint, a forbidden phrase, a required output structure, or a clarifying example. Diff should be < 30 added/removed lines. |
 | **`output-format`** / **`quality-regression`** | Re-read the target skill's own output spec in its `SKILL.md`. Edit the skill so the next run satisfies that spec. Cite the exact requirement (section / line) in the PR body. |