cursor-route 0.1.13 → 0.1.15

package/CHANGELOG.md

@@ -4,6 +4,18 @@
 
 _(none)_
 
+## 0.1.15 — 2026-09-11
+
+- Mid default is `deepseek-flash` (DeepSeek V4.1 Flash, native vision). `--model pro` and `--model vision` are legacy aliases to the same id. Pro is off rotation.
+- `--model flash|pro|vision` and legacy full ids (`deepseek-flash`, `deepseek-v4-flash`, `deepseek-v4.1-flash`, `deepseek-v4-pro`, `deepseek-v4-pro[1m]`, `deepseek-v4-flash-vision`, `deepseek-v4-flash-vision-exp`) all launch `deepseek-flash`. `job.model` still records the alias the operator passed.
+- Mid lane still `claude-ds`; no worker swap. Hero demo fixture regenerated to **0.1.15**.
+
+## 0.1.14 — 2026-09-02
+
+- Easy-lane OpenRouter live pick now **tier-ranks** free models like the toolkit: Qwen 100, GLM/Kimi 95, DeepSeek 90, coding/chat families 70, **Nemotron 15**, default 40. Score adds a context bonus capped at 131072 (`tier + min(ctx, 131072)/131072 × 5`) so a huge Nemotron 550B free model can no longer outrank Qwen/GLM/Kimi on context alone.
+- Tests: all seven tier buckets; equal-score tier/id tiebreakers; same-tier 131072-vs-262144 cap; ranked-pick cache round-trip.
+- Mid lane still `claude-ds`; no worker swap. Hero demo fixture regenerated to **0.1.14**.
+
 ## 0.1.13 — 2026-08-29
 
 - OpenRouter live-pick: do **not** cache the fetch-fail fallback (`openrouter/free`) — health no longer presents a stale fallback as `now openrouter/free`. Health says `cached <id>` only for a ranked catalog hit; otherwise `no catalog cache, fetch-fail fallback openrouter/free`.

package/README.md

@@ -116,17 +116,17 @@ docs/fixtures/generate-hero-demo.sh      # regenerate docs/fixtures/hero-demo.lo
 
 ```text
 $ cursor-route --version
-0.1.13
+0.1.15
 
 $ CURSOR_ROUTE_RELAXED=1 cursor-route health
-cursor-route v0.1.13
+cursor-route v0.1.15
 health: OK
 
 $ cursor-route start --lane mid --model flash --dry-run "Add a unit test for shellQuote"
 dry-run job a1b2c3d4
 worker: claude-ds
 model:  flash
-command: cd '~/Projects/cursor-route' && '~/.local/bin/claude-ds' -PromptFile '~/.local/share/cursor-route/jobs/a1b2c3d4.prompt' -Model 'deepseek-v4-flash' --dangerously-skip-permissions
+command: cd '~/Projects/cursor-route' && '~/.local/bin/claude-ds' -PromptFile '~/.local/share/cursor-route/jobs/a1b2c3d4.prompt' -Model 'deepseek-flash' --dangerously-skip-permissions
 
 $ cursor-route jobs --json
 []
@@ -145,26 +145,27 @@ A real hero GIF is still pending — recording steps live in [docs/DEMO_GIF.md](
 
 Always-approve is **on** by default for **coding worktrees only**. It does not authorize LIVE Discord, trading, or irreversible SaaS. Opt out: `--ask` or `CURSOR_ROUTE_ASK=1`.
 
-### Mid models (Flash vs Pro)
+### Mid models (Flash)
 
 | Flag | Model id | When |
 |------|----------|------|
-| `--model flash` (default) | `deepseek-v4-flash` | Cheap mid execute. Prefer this when Grok **usage** is out |
-| `--model vision` | `deepseek-v4-flash-vision-exp` | Screenshots / ui mocks / image prompts (or auto-pick) |
-| `--model pro` | `deepseek-v4-pro` | Harder mid / **hard backup** only — not the default Grok-out stand-in |
-| `--model deepseek-v4-pro[1m]` | `deepseek-v4-pro[1m]` | Large-context Pro (SKU preserved) |
+| `--model flash` (default) | `deepseek-flash` | Cheap mid execute. Prefer this when Grok **usage** is out |
+| `--model vision` | `deepseek-flash` | Screenshots / ui mocks / image prompts (or auto-pick). V4.1 Flash has native vision |
+| `--model pro` | `deepseek-flash` | Legacy alias — Pro is off rotation |
+
+This map matches DeepSeek's 2026-09-10 API id `deepseek-flash`. `--model pro` and `--model vision` are legacy aliases to the same id (including `deepseek-v4-pro[1m]`).
 
 ```bash
 cursor-route start --lane mid "…"                    # Flash
-cursor-route start --lane mid --model vision "…"     # Vision Flash
-cursor-route start --lane mid --model pro "…"        # Pro (harder mid / hard backup)
+cursor-route start --lane mid --model vision "…"     # legacy alias → deepseek-flash
+cursor-route start --lane mid --model pro "…"        # legacy alias → deepseek-flash
 # Or set default without a flag:
 export CURSOR_ROUTE_DS_MODEL=flash   # also honors ANTHROPIC_MODEL; --model overrides
 ```
 
-If `--model` and `CURSOR_ROUTE_DS_MODEL` / `ANTHROPIC_MODEL` are unset, a prompt that looks like a screenshot/image/png/jpg/jpeg/webp/ui mock/multimodal/vision auto-picks vision Flash. Explicit `--model flash|pro|vision` always wins.
+If `--model` and `CURSOR_ROUTE_DS_MODEL` / `ANTHROPIC_MODEL` are unset, a prompt that looks like a screenshot/image/png/jpg/jpeg/webp/ui mock/multimodal/vision auto-picks the vision alias (`job.model` is `vision`; launch id is still `deepseek-flash`). Explicit `--model flash|pro|vision` always wins.
 
-**Grok auth ≠ usage-out:** if `cursor-route health` shows `worker:grok` ✗, run `grok login` (or set `XAI_API_KEY`). That is auth, not the Pro case. When Grok **usage** is exhausted, stay on `--lane mid --model flash` (cheap default). `--model pro` is harder mid / hard backup only.
+**Grok auth ≠ usage-out:** if `cursor-route health` shows `worker:grok` ✗, run `grok login` (or set `XAI_API_KEY`). That is auth, not a model switch. When Grok **usage** is exhausted, stay on `--lane mid --model flash` (cheap default). `--model pro` is a legacy alias to `deepseek-flash`.
 
 `CURSOR_ROUTE_ALLOW_ANTHROPIC=1` is an expensive escape hatch: it does **not** pass DeepSeek `--model` ids (Anthropic would reject them).
 
@@ -180,18 +181,18 @@ npm i -g @anthropic-ai/claude-code
 export ANTHROPIC_BASE_URL=https://api.deepseek.com/anthropic
 export ANTHROPIC_AUTH_TOKEN=YOUR_DEEPSEEK_API_KEY   # from platform.deepseek.com
 # Optional shell defaults (CLI --model overrides ANTHROPIC_MODEL for the job):
-export ANTHROPIC_DEFAULT_HAIKU_MODEL=deepseek-v4-flash
-export CLAUDE_CODE_SUBAGENT_MODEL=deepseek-v4-flash
+export ANTHROPIC_DEFAULT_HAIKU_MODEL=deepseek-flash
+export CLAUDE_CODE_SUBAGENT_MODEL=deepseek-flash
 
 cursor-route health          # worker:claude-ds should be ✓; `lane:mid` is ✓ only when DeepSeek is proven
 cursor-route start --lane mid "…"                 # Flash (default; also when Grok usage is out)
-cursor-route start --lane mid --model vision "…"  # screenshots / ui mocks
-cursor-route start --lane mid --model pro "…"     # harder mid / hard backup only
+cursor-route start --lane mid --model vision "…"  # screenshots / ui mocks (same launch id)
+cursor-route start --lane mid --model pro "…"     # legacy alias → deepseek-flash
 ```
 
 Persist the same vars under `~/.claude/settings.json` → `"env": { … }` if you want them every shell.
 
-**Also accepted:** `claude-ds` or `deepseek-claude` on PATH (Cemini shims). The adapter passes `-Model deepseek-v4-flash|deepseek-v4-pro`.
+**Also accepted:** `claude-ds` or `deepseek-claude` on PATH (Cemini shims). The adapter passes `-Model deepseek-flash`.
 
 **Experimental:** `--worker deepseek` runs the official DeepSeek Harness (`dsh`, npm `@deepseek-ai/dsh`) as an opt-in worker — **mid stays on `claude-ds`**; don't default lanes here.
 
@@ -201,10 +202,10 @@ export DEEPSEEK_API_KEY=YOUR_DEEPSEEK_API_KEY   # from platform.deepseek.com
 
 cursor-route health                              # worker:deepseek should be ✓
 cursor-route start --worker deepseek "…"         # Flash (default)
-cursor-route start --worker deepseek --model pro "…"   # harder mid / hard backup only
+cursor-route start --worker deepseek --model pro "…"   # legacy alias → deepseek-flash
 ```
 
-The adapter launches `dsh --profile headless` with a **per-job Cordis patch** (`jobs/<id>.dsh-patch.yml`, mode 0600) that pins `--model flash|pro|vision` (`deepseek-v4-pro[1m]` preserved) — it never rewrites `~/.dsh/settings.yaml`, so parallel jobs don't race. Always-approve maps to `DSH_PERMISSION_MODE=danger-full-access`; `--ask` drops to `workspace-write`. Your `DEEPSEEK_API_KEY` travels via env only — never in the launch command or patch. Override the binary with `CURSOR_ROUTE_DSH_BIN`.
+The adapter launches `dsh --profile headless` with a **per-job Cordis patch** (`jobs/<id>.dsh-patch.yml`, mode 0600) that pins `--model flash|pro|vision` to `deepseek-flash` — it never rewrites `~/.dsh/settings.yaml`, so parallel jobs don't race. Always-approve maps to `DSH_PERMISSION_MODE=danger-full-access`; `--ask` drops to `workspace-write`. Your `DEEPSEEK_API_KEY` travels via env only — never in the launch command or patch. Override the binary with `CURSOR_ROUTE_DSH_BIN`.
 
 **Not the default:** bare `claude` still talking to Anthropic. Health refuses that so a misconfigured install cannot silently burn frontier $ rates. Escape hatch only: `CURSOR_ROUTE_ALLOW_ANTHROPIC=1`.
 
@@ -214,7 +215,9 @@ No DeepSeek yet? Use `--lane hard` / `--worker grok` (X Premium).
 
 `--lane easy` / `--worker openrouter` sends wording/draft prompts to OpenRouter
 and **live-picks the best free text model** at request time (`GET /models`, rank
-`:free` or $0 text models). Do not hardcode a specific model id as the default.
+`:free` or $0 text models). The live rank prefers Qwen/GLM/Kimi coding free
+models over huge general free models (e.g. Nemotron 550B). Do not hardcode a
+specific model id as the default.
 Pin with `CURSOR_ROUTE_OPENROUTER_MODEL` or `--model provider/model`. Empty /
 `free` = live pick. If the catalog fetch fails, the fallback is OpenRouter's
 **router** id `openrouter/free` (a live router, not a locked model). Get a key

package/dist/adapters/claude-ds.js

@@ -128,8 +128,16 @@ export function midDeepSeekProofDetail() {
 }
 function pickModel(requested, modelId) {
     if (modelId) {
-        const alias = requested ?? resolveDsModel(modelId).alias;
-        return { alias, id: modelId };
+        try {
+            const resolved = resolveDsModel(modelId);
+            return { alias: requested ?? resolved.alias, id: resolved.id };
+        }
+        catch (e) {
+            // Unknown id with an explicit alias: keep the raw id so dsh YAML checks can still reject it.
+            if (requested)
+                return { alias: requested, id: modelId };
+            throw e;
+        }
     }
     if (requested) {
         return { alias: requested, id: DS_MODEL_IDS[requested] };

package/dist/adapters/deepseek.js

@@ -32,8 +32,16 @@ function findDsh() {
 /** Same resolution as claude-ds: passed model/modelId wins, else env, else Flash. */
 function pickModel(requested, modelId) {
     if (modelId) {
-        const alias = requested ?? resolveDsModel(modelId).alias;
-        return { alias, id: modelId };
+        try {
+            const resolved = resolveDsModel(modelId);
+            return { alias: requested ?? resolved.alias, id: resolved.id };
+        }
+        catch (e) {
+            // Unknown id with an explicit alias: keep the raw id so YAML checks can still reject it.
+            if (requested)
+                return { alias: requested, id: modelId };
+            throw e;
+        }
     }
     if (requested) {
         return { alias: requested, id: DS_MODEL_IDS[requested] };

package/dist/cli.js

@@ -31,7 +31,7 @@ Usage:
 Start options:
   --worker <grok|claude-ds|openrouter|deepseek|opencode>  Worker adapter (default: grok; deepseek/opencode = opt-in)
   --lane <easy|mid|hard>                Lane → worker (easy=openrouter, mid=claude-ds, hard=grok)
-  --model <flash|pro|vision|free|provider/model>  claude-ds/deepseek: flash|pro|vision. opencode/openrouter: free (live catalog pick) or provider/model
+  --model <flash|pro|vision|free|provider/model>  claude-ds/deepseek: flash|pro|vision (aliases map to deepseek-flash). opencode/openrouter: free (live catalog pick) or provider/model
   --dir <path>                          Working directory (default: cwd)
   --ask                                 Disable always-approve for this job
   --dry-run                             Print launch command; do not start
@@ -44,7 +44,7 @@ Env:
   CURSOR_ROUTE_MAX_JOBS              Max active jobs (default: 50)
   CURSOR_ROUTE_RELAXED=1             health OK without tmux/workers (CI / infra smoke)
   CURSOR_ROUTE_ALLOW_ANTHROPIC=1     Allow mid-lane on Anthropic Claude (expensive; not default; --model ignored)
-  CURSOR_ROUTE_DS_MODEL              Default mid model flash|pro|vision (or full ids / deepseek-v4-pro[1m]); overridden by --model
+  CURSOR_ROUTE_DS_MODEL              Default mid model flash|pro|vision (aliases map to deepseek-flash); overridden by --model
   CURSOR_ROUTE_GROK_BIN              Override the grok binary path (tests / power users)
   CURSOR_ROUTE_CLAUDE_DS_BIN         Override the claude-ds binary path (tests / power users)
   CURSOR_ROUTE_DSH_BIN               Override the dsh binary path (tests / power users)

package/dist/config.js

@@ -4,10 +4,12 @@ import { defaultJobsDir } from "./runtime.js";
 export const WORKERS = ["grok", "claude-ds", "openrouter", "deepseek", "opencode"];
 export const LANES = ["easy", "mid", "hard"];
 export const DS_MODELS = ["flash", "pro", "vision"];
+/** DeepSeek V4.1 Flash (native vision). Matches the 2026-09-10 API id. */
+export const DS_FLASH_ID = "deepseek-flash";
 export const DS_MODEL_IDS = {
-    flash: "deepseek-v4-flash",
-    pro: "deepseek-v4-pro",
-    vision: "deepseek-v4-flash-vision-exp",
+    flash: DS_FLASH_ID,
+    pro: DS_FLASH_ID,
+    vision: DS_FLASH_ID,
 };
 /** Prompt looks like a screenshot / image / ui mock — vision auto-pick. */
 const VISION_PROMPT_RE = /screenshot|\bimage\b|\.png|\.jpe?g|\.webp|ui mock|multimodal|\bvision\b/i;
@@ -15,26 +17,31 @@ export function promptLooksLikeVision(prompt) {
     return VISION_PROMPT_RE.test(prompt);
 }
 /**
- * Resolve --model / env to alias + concrete model id.
- * Preserves `deepseek-v4-pro[1m]` (does not silently strip the SKU).
+ * Resolve --model / env to alias + launch id.
+ * V4.1 Flash is native vision; Pro is a legacy alias, not a rotation step.
+ * All known aliases launch `deepseek-flash`.
  * Empty → Flash.
  */
 export function resolveDsModel(raw) {
     if (!raw || !raw.trim()) {
-        return { alias: "flash", id: DS_MODEL_IDS.flash };
+        return { alias: "flash", id: DS_FLASH_ID };
     }
-    const v = raw.trim().toLowerCase();
-    if (v === "flash" || v === "deepseek-v4-flash") {
-        return { alias: "flash", id: DS_MODEL_IDS.flash };
+    // Trim, lower-case, strip a trailing `[…]` SKU (`deepseek-v4-pro[1m]` → `deepseek-v4-pro`).
+    const v = raw.trim().toLowerCase().replace(/\[[^\]]*\]\s*$/, "");
+    if (v === "flash" ||
+        v === "deepseek-flash" ||
+        v === "deepseek-v4-flash" ||
+        v === "deepseek-v4.1-flash") {
+        return { alias: "flash", id: DS_FLASH_ID };
     }
     if (v === "pro" || v === "deepseek-v4-pro") {
-        return { alias: "pro", id: DS_MODEL_IDS.pro };
+        return { alias: "pro", id: DS_FLASH_ID };
     }
-    if (v === "vision" || v === "deepseek-v4-flash-vision-exp") {
-        return { alias: "vision", id: DS_MODEL_IDS.vision };
-    }
-    if (v === "deepseek-v4-pro[1m]") {
-        return { alias: "pro", id: "deepseek-v4-pro[1m]" };
+    if (v === "vision" ||
+        v === "flash-vision" ||
+        v === "deepseek-v4-flash-vision" ||
+        v === "deepseek-v4-flash-vision-exp") {
+        return { alias: "vision", id: DS_FLASH_ID };
     }
     throw new Error(`Invalid DeepSeek model ${raw}; expected flash|pro|vision`);
 }
@@ -72,7 +79,7 @@ function maxConcurrentJobsFromEnv() {
  */
 export const config = {
     product: "cursor-route",
-    version: "0.1.13",
+    version: "0.1.15",
     get jobsDir() {
         return defaultJobsDir();
     },

package/dist/or-free.js

@@ -12,7 +12,6 @@ export const OPENROUTER_MODELS_URL_DEFAULT = "https://openrouter.ai/api/v1/model
 /** Fetch-fail / empty-rank fallback — OpenRouter live router, not a locked model. */
 export const OPENROUTER_FALLBACK_MODEL = "openrouter/free";
 const EXCLUDE_RE = /lyria|whisper|tts|embed|embedding|image|vision-only|audio|diffusion|flux|stable-diffusion|moderation/i;
-const BOOST_RE = /coder|instruct|chat|nemotron|qwen|llama|gemma|gpt-oss|kimi|glm|deepseek/i;
 function orModelsUrl() {
     const base = (process.env.OPENROUTER_BASE_URL || "https://openrouter.ai/api/v1")
         .trim()
@@ -59,9 +58,28 @@ export function isOrFreeModel(m) {
     const tagged = id.endsWith(":free");
     return tagged || pricedFree;
 }
-/** Higher boost wins. Coding/chat families beat generic free. No hardcoded id. */
+/**
+ * Toolkit-parity tier for a live OpenRouter free model id (first match wins).
+ * Huge general free models (Nemotron) rank well below coding/chat families so
+ * the easy-lane live pick does not land on a slow 550B generalist. Order
+ * matters: `qwen` is checked before `nemotron`, and the generic coding bucket
+ * (70) sits before `nemotron` exactly like the toolkit, so
+ * `nvidia/nemotron-3-550b:free` scores 15 — never 70. No hardcoded model id.
+ */
 export function orFreeBoost(id) {
-    return BOOST_RE.test(id) ? 20 : 10;
+    if (/qwen/i.test(id))
+        return 100;
+    if (/(^|\/)z-ai\/glm|glm/i.test(id))
+        return 95;
+    if (/kimi|moonshot/i.test(id))
+        return 95;
+    if (/deepseek|hy-/i.test(id))
+        return 90;
+    if (/coder|instruct|chat|llama|gemma|gpt-oss|minimax/i.test(id))
+        return 70;
+    if (/nemotron/i.test(id))
+        return 15;
+    return 40;
 }
 export function rankOrFreeModels(models) {
     const out = [];
@@ -77,18 +95,24 @@ export function rankOrFreeModels(models) {
             continue;
         }
         const ctx = Number.isFinite(Number(m.context_length)) ? Number(m.context_length) : 0;
+        const tier = orFreeBoost(id);
+        // Cap the context bonus at 131072 so a huge Nemotron ctx cannot outrank a
+        // higher tier on context alone: score = tier + min(ctx, 131072)/131072 * 5.
+        const score = tier + (Math.min(ctx, 131_072) / 131_072) * 5;
         out.push({
             id,
             name: (m.name || raw).trim(),
             context_length: ctx,
-            boost: orFreeBoost(id),
+            boost: score,
         });
     }
+    // Sort like the toolkit: score desc, then tier desc, then id asc.
     out.sort((a, b) => {
         if (b.boost !== a.boost)
             return b.boost - a.boost;
-        if (b.context_length !== a.context_length)
-            return b.context_length - a.context_length;
+        const tierDiff = orFreeBoost(b.id) - orFreeBoost(a.id);
+        if (tierDiff !== 0)
+            return tierDiff;
         return a.id.localeCompare(b.id);
     });
     return out;

package/docs/briefs/WORKING.md

@@ -1,9 +1,9 @@
 ---
 title: cursor-route workspace — working brief (edit in place)
 repo: ~/Projects/cursor-route
-npm: cursor-route@0.1.13 (publishing this slice)
+npm: cursor-route@0.1.15
 created: 2026-08-12
-updated: 2026-08-29
+updated: 2026-09-11
 ---
 
 # cursor-route — living brief
@@ -19,13 +19,13 @@ Cursor Agent plans. Workers run in tmux via `cursor-route`:
 | Lane | Worker | Intent |
 |------|--------|--------|
 | `easy` | OpenRouter free (live pick) | Wording / drafts — non-secret prompts only |
-| `mid` | claude-ds (DeepSeek behind Claude Code) | Default implement (**Flash**; `--model vision` for screenshots; `--model pro` harder mid / hard backup only) |
+| `mid` | claude-ds (DeepSeek behind Claude Code) | Default implement (**Flash** `deepseek-flash`; `--model vision` / `--model pro` are legacy aliases to the same id) |
 | `hard` | Grok CLI | Hard implement |
 | opt-in | OpenCode | `--worker opencode` coding agent; `--model free` ranks live Zen catalog (Ox Alpha first while listed) |
 
 Always-approve on for coding worktrees (`--ask` / `CURSOR_ROUTE_ASK=1` to opt out) — not LIVE Discord/trading. Jobs live in `~/.local/share/cursor-route/jobs`, not in this clone.
 
-Install: `npm i -g cursor-route` → **0.1.13**. Release notes: [CHANGELOG.md](../../CHANGELOG.md).
+Install: `npm i -g cursor-route` → **0.1.15** (not LIVE until parent publishes). Release notes: [CHANGELOG.md](../../CHANGELOG.md).
 
 ## Open (edit / check off)
 
@@ -43,6 +43,8 @@ Install: `npm i -g cursor-route` → **0.1.13**. Release notes: [CHANGELOG.md](.
 - [x] **Live Zen free pick** — 0.1.11: `--model free` ranks `GET https://opencode.ai/zen/v1/models` (Ox Alpha first while listed; ~15 min cache; Ox Alpha fallback if fetch fails). Pin with `CURSOR_ROUTE_OPENCODE_MODEL`. Health stays offline (cache/fallback only).
 - [x] **Flash-first + vision + live OpenRouter pick** — 0.1.12: `--model flash` stays the cheap Grok-out default; `--model vision` + screenshot auto-pick; easy lane live-picks free OpenRouter text models (`openrouter/free` is fetch-fail fallback only). Health never fetches `/models`. `route-orch` ProgRouter + MoRe one-liners. Hero fixture regenerated to 0.1.12. No npm publish in this slice.
 - [x] **Kimi 0.1.13 follow-up** — do not cache OpenRouter fallback; health labels uncached fallback; no Authorization on GET /models; env-invalid model error names the env vars; `\bimage\b` vision trigger.
+- [x] **Super-audit OR ranker parity (0.1.14)** — tier-rank Qwen/GLM/Kimi above Nemotron 550B in `or-free.ts` (match agent-toolkit `select-openrouter-free-model.ps1`); brief `docs/briefs/2026-09-02_super-audit-or-ranker-parity.md`
+- [x] **V4.1 Flash wire (0.1.15)** — mid default `deepseek-flash` (native vision). `--model pro` / `--model vision` are legacy aliases to the same id. Pro is off rotation. Mid still `claude-ds`. Hero fixture regenerated. npm not LIVE until parent publishes. Brief `docs/briefs/2026-09-11_v41-flash-not-pro.md`
 - [ ] **Hero GIF** — still outstanding; dry-run fixture ships as the substitute for now (`docs/fixtures/hero-demo.log` — see `docs/DEMO_GIF.md`)
 - [x] **Do not** paste private `ROUTE_KIT`, SIP, prod paths, or hang-watchdog env into this public repo
 
@@ -79,4 +81,7 @@ Install: `npm i -g cursor-route` → **0.1.13**. Release notes: [CHANGELOG.md](.
 | 2026-08-21 | Opt-in `--worker opencode` (Zen free `opencode/big-pickle`, `--auto`, no config rewrite); mid stays claude-ds → 0.1.10 LIVE. |
 | 2026-08-21 | Live Zen free pick (Ox Alpha first while listed; OpenRouter-style catalog rank) → 0.1.11 LIVE. |
 | 2026-08-29 | Flash-first docs, `--model vision` + auto-pick, live OpenRouter free pick, route-orch K318/K322, hero fixture 0.1.12. No npm publish. |
-| 2026-08-29 | Kimi audit follow-up → 0.1.13: no fallback cache, health labels, unauth GET /models, env error wording. |
+| 2026-08-29 | Kimi audit follow-up → 0.1.13 LIVE on npm: no fallback cache, health labels, unauth GET /models, env error wording. |
+| 2026-09-02 | Tier-rank OpenRouter free picks in `or-free.ts` (Qwen 100 / GLM-Kimi 95 / DeepSeek 90 / coding 70 / Nemotron 15 / default 40; ctx bonus capped at 131072) → 0.1.14. Commit + tag `v0.1.14`. Mid stays claude-ds. |
+| 2026-09-02 | Kimi audit follow-up → ranker tests (seven tiers, score/id ties, 131072 cap, cache round-trip) then **0.1.14 LIVE** on npm. |
+| 2026-09-11 | Map mid models to `deepseek-flash` (V4.1 Flash, native vision). `--model pro` / `--model vision` are legacy aliases. Pro is off rotation. Mid still `claude-ds` → 0.1.15. |

package/docs/demo-notes.md

@@ -6,10 +6,10 @@ Simulated capture output for README / tweet assets — replace with a real GIF o
 
 ```
 $ cursor-route --version
-0.1.13
+0.1.15
 
 $ cursor-route health
-cursor-route v0.1.13
+cursor-route v0.1.15
 health: OK
   ✓ tmux
   ✓ runtime          bun ok
@@ -32,7 +32,7 @@ $ cursor-route jobs --json
 
 Verified locally (2026-08-10): headless `claude-ds` smoke returned `CURSOR_ROUTE_SMOKE_OK`.
 Grok smoke hit 402 (Build usage balance exhausted) — auth/PATH wiring works; top up Grok Build for live demos.
-Use `--model flash` when Grok **usage** is out (cheap default). `--model pro` is harder mid / hard backup only — not the default Grok-out stand-in. A missing `grok login` is auth, not the Pro case.
+Use `--model flash` when Grok **usage** is out (cheap default). `--model pro` and `--model vision` are legacy aliases to `deepseek-flash`. A missing `grok login` is auth, not a model switch.
 
 Current commands: `health`, `start`, `jobs`, `status`, `capture`, `send`, `attach`, `kill`, `sessions`, `clean`.
 Headless demos (no tmux) use `--no-tmux` and `capture`/`status` instead of `attach`/`send`.

package/docs/fixtures/hero-demo.log

@@ -1,15 +1,15 @@
 $ cursor-route --version
-0.1.13
+0.1.15
 
 $ CURSOR_ROUTE_RELAXED=1 cursor-route health
-cursor-route v0.1.13
+cursor-route v0.1.15
 health: OK
 
   ✓ tmux             ok
   ✓ runtime          bun ok
   ✓ script(1)        ok (tty log capture)
   ✓ worker:grok      ok (auth checked at first start — run grok login if jobs fail) @ ~/.grok/bin/grok
-  ✓ worker:claude-ds ok (claude-ds (DeepSeek shim); default model deepseek-v4-flash) @ ~/.local/bin/claude-ds
+  ✓ worker:claude-ds ok (claude-ds (DeepSeek shim); default model deepseek-flash) @ ~/.local/bin/claude-ds
   ✗ worker:openrouter OPENROUTER_API_KEY not set — export your OpenRouter key (easy lane live-picks a free model at start; pin with CURSOR_ROUTE_OPENROUTER_MODEL) @ node '~/Projects/cursor-route/dist/openrouter-run.js'
   ✗ worker:deepseek  dsh (@deepseek-ai/dsh) not found — install: npm i -g @deepseek-ai/dsh. Mid default remains claude-ds.
   ✓ worker:opencode  ok (opencode run; --model free = live Zen pick, now opencode/x-preview-f-free; auth at first start — opencode auth login if jobs fail; mid default remains claude-ds) @ ~/.local/bin/opencode
@@ -22,7 +22,7 @@ $ cursor-route start --lane mid --model flash --dry-run "Add a unit test for she
 dry-run job a1b2c3d4
 worker: claude-ds
 model:  flash
-command: cd '~/Projects/cursor-route' && '~/.local/bin/claude-ds' -PromptFile '~/.local/share/cursor-route/jobs/a1b2c3d4.prompt' -Model 'deepseek-v4-flash' --dangerously-skip-permissions
+command: cd '~/Projects/cursor-route' && '~/.local/bin/claude-ds' -PromptFile '~/.local/share/cursor-route/jobs/a1b2c3d4.prompt' -Model 'deepseek-flash' --dangerously-skip-permissions
 
 $ cursor-route start --lane easy --dry-run "Rewrite this FAQ answer in 3 sentences"
 dry-run job b2c3d4e5
@@ -48,7 +48,7 @@ $ cursor-route start --lane mid --dry-run --json "Add a failing test then make i
   "alwaysApprove": true,
   "tmuxSession": "cursor-route-d4e5f6a7",
   "createdAt": "2026-08-13T00:00:00.000Z",
-  "command": "cd '~/Projects/cursor-route' && '~/.local/bin/claude-ds' -PromptFile '~/.local/share/cursor-route/jobs/d4e5f6a7.prompt' -Model 'deepseek-v4-flash' --dangerously-skip-permissions"
+  "command": "cd '~/Projects/cursor-route' && '~/.local/bin/claude-ds' -PromptFile '~/.local/share/cursor-route/jobs/d4e5f6a7.prompt' -Model 'deepseek-flash' --dangerously-skip-permissions"
 }
 
 $ cursor-route jobs --json

package/llms.txt

@@ -2,7 +2,7 @@
 
 > Cursor stays the planner. DeepSeek (mid), Grok CLI (hard), and OpenRouter free models (easy) run parallel coding workers in tmux. `--worker opencode` is an opt-in coding agent on OpenCode Zen free models.
 
-MIT CLI + Cursor skill. npm: https://www.npmjs.com/package/cursor-route (latest **0.1.13**)
+MIT CLI + Cursor skill. npm: https://www.npmjs.com/package/cursor-route (latest **0.1.15**)
 GitHub: https://github.com/cemini23/cursor-route
 
 ## FAQ
@@ -14,15 +14,16 @@ cursor-route is a public MIT CLI and Cursor skill that runs parallel coding work
 It uses the familiar strategist and worker-pane shape, but it is not a Codex clone. cursor-route uses Cursor as the planner and DeepSeek plus Grok CLI as workers. Codex is not required.
 
 ### Does mid lane use Anthropic Claude?
-No. The mid worker is DeepSeek. Claude Code is the harness, configured with `ANTHROPIC_BASE_URL=https://api.deepseek.com/anthropic` and a DeepSeek key in `ANTHROPIC_AUTH_TOKEN`. Default model is Flash (`--model flash`) — keep Flash when Grok **usage** is out. `--model vision` is for screenshots / ui mocks. `--model pro` is harder mid / **hard backup only**, not the default Grok-out stand-in. A missing `grok login` is auth, not the Pro case.
+No. The mid worker is DeepSeek. Claude Code is the harness, configured with `ANTHROPIC_BASE_URL=https://api.deepseek.com/anthropic` and a DeepSeek key in `ANTHROPIC_AUTH_TOKEN`. Default model is Flash (`--model flash` → `deepseek-flash`) — keep Flash when Grok **usage** is out. `--model vision` and `--model pro` are legacy aliases to the same id. A missing `grok login` is auth, not a model switch.
 
 ### Flash vs Pro?
 | Flag | Model | When |
 |------|-------|------|
-| `--model flash` (default) | `deepseek-v4-flash` | Cheap mid execute. Prefer this when Grok **usage** is out |
-| `--model vision` | `deepseek-v4-flash-vision-exp` | Screenshots / ui mocks / image prompts (or auto-pick) |
-| `--model pro` | `deepseek-v4-pro` | Harder mid / **hard backup** only — not the default Grok-out stand-in |
-| `--model deepseek-v4-pro[1m]` | preserved SKU | Large-context Pro |
+| `--model flash` (default) | `deepseek-flash` | Cheap mid execute. Prefer this when Grok **usage** is out |
+| `--model vision` | `deepseek-flash` | Screenshots / ui mocks / image prompts (or auto-pick). V4.1 Flash has native vision |
+| `--model pro` | `deepseek-flash` | Legacy alias — Pro is off rotation |
+
+This map matches DeepSeek's 2026-09-10 API id `deepseek-flash`.
 
 ### How do I use OpenCode free models?
 Install OpenCode (`npm i -g opencode-ai`), run `opencode auth login`, then `cursor-route start --worker opencode "…"`. `--model free` ranks live Zen free models (Ox Alpha first while listed). Mid stays `claude-ds`.
@@ -37,6 +38,6 @@ The cursor-route code is open source under MIT. It does not make the worker serv
 
 - npm: `npm i -g cursor-route`
 - Health: `cursor-route health`
-- Mid: `cursor-route start --lane mid "…"` (Flash; keep Flash when Grok usage is out) · `--model vision` for screenshots · `--model pro` harder mid / hard backup only
+- Mid: `cursor-route start --lane mid "…"` (Flash `deepseek-flash`; keep Flash when Grok usage is out) · `--model vision` / `--model pro` are legacy aliases to the same id
 - OpenCode (opt-in): `cursor-route start --worker opencode "…"`
 - Skill: `/route-orch` (Cursor)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "cursor-route",
-  "version": "0.1.13",
+  "version": "0.1.15",
   "description": "Cursor stays the brain. Grok CLI + DeepSeek (claude-ds) + OpenRouter easy lane + OpenCode (opt-in free) are the parallel army \u2014 lane-aware /route orchestration in tmux.",
   "type": "module",
   "license": "MIT",

package/skills/route-orch/SKILL.md

@@ -28,7 +28,7 @@ You are the **orchestrator**. Do **not** implement bulk code in this Cursor sess
 | `mid` | `claude-ds` (DeepSeek via Claude Code harness) | Standard implement / refactor |
 | `hard` | `grok` | Premium plan in Cursor → Grok implement |
 
-Free OpenRouter models may log prompts — keep secrets off the easy lane (the CLI refuse gate still applies). Easy lane **live-picks** the best free text model at request time (unset / `--model free`). Pin with `CURSOR_ROUTE_OPENROUTER_MODEL` or `--model provider/model`. Fetch-fail fallback is the OpenRouter router `openrouter/free` only.
+Free OpenRouter models may log prompts — keep secrets off the easy lane (the CLI refuse gate still applies). Easy lane **live-picks** the best free text model at request time (unset / `--model free`), preferring Qwen/GLM/Kimi coding free models over huge general free models (Nemotron). Pin with `CURSOR_ROUTE_OPENROUTER_MODEL` or `--model provider/model`. Fetch-fail fallback is the OpenRouter router `openrouter/free` only.
 
 ## claude-ds models
 
@@ -36,10 +36,11 @@ One harness. Do not install a second coding loop.
 
 | Flag | Model id | When |
 |------|----------|------|
-| `--model flash` (default) | `deepseek-v4-flash` | Cheap mid execute. Prefer this when Grok **usage** is out |
-| `--model vision` | `deepseek-v4-flash-vision-exp` | Screenshots / ui mocks / image prompts (or auto-pick) |
-| `--model pro` | `deepseek-v4-pro` | Harder mid / **hard backup** only — not the default Grok-out stand-in |
-| `--model deepseek-v4-pro[1m]` | preserved SKU | Large-context Pro |
+| `--model flash` (default) | `deepseek-flash` | Cheap mid execute. Prefer this when Grok **usage** is out |
+| `--model vision` | `deepseek-flash` | Screenshots / ui mocks / image prompts (or auto-pick) |
+| `--model pro` | `deepseek-flash` | Legacy alias — Pro is off rotation |
+
+V4.1 Flash has native vision; `--model pro` is a legacy alias (off rotation).
 
 ```bash
 cursor-route start --lane mid --dir "$PWD" "…"
@@ -47,7 +48,7 @@ cursor-route start --lane mid --model vision --dir "$PWD" "…"
 cursor-route start --lane mid --model pro --dir "$PWD" "…"
 ```
 
-If `worker:grok` is ✗ on health, that is usually **auth** (`grok login` / `XAI_API_KEY`) — not the Pro case. When Grok **usage** is out, stay on Flash.
+If `worker:grok` is ✗ on health, that is usually **auth** (`grok login` / `XAI_API_KEY`) — not a model switch. When Grok **usage** is out, stay on Flash.
 
 ## Experimental: --worker deepseek (dsh)
 

package/src/adapters/claude-ds.ts

@@ -145,8 +145,14 @@ export function midDeepSeekProofDetail(): string {
 
 function pickModel(requested?: DsModelAlias, modelId?: string): { alias: DsModelAlias; id: string } {
   if (modelId) {
-    const alias = requested ?? resolveDsModel(modelId).alias;
-    return { alias, id: modelId };
+    try {
+      const resolved = resolveDsModel(modelId);
+      return { alias: requested ?? resolved.alias, id: resolved.id };
+    } catch (e) {
+      // Unknown id with an explicit alias: keep the raw id so dsh YAML checks can still reject it.
+      if (requested) return { alias: requested, id: modelId };
+      throw e;
+    }
   }
   if (requested) {
     return { alias: requested, id: DS_MODEL_IDS[requested] };

package/src/adapters/deepseek.ts

@@ -43,8 +43,14 @@ function pickModel(
   modelId?: string,
 ): { alias: DsModelAlias; id: string } {
   if (modelId) {
-    const alias = requested ?? resolveDsModel(modelId).alias;
-    return { alias, id: modelId };
+    try {
+      const resolved = resolveDsModel(modelId);
+      return { alias: requested ?? resolved.alias, id: resolved.id };
+    } catch (e) {
+      // Unknown id with an explicit alias: keep the raw id so YAML checks can still reject it.
+      if (requested) return { alias: requested, id: modelId };
+      throw e;
+    }
   }
   if (requested) {
     return { alias: requested, id: DS_MODEL_IDS[requested] };

package/src/cli.test.ts

@@ -52,24 +52,32 @@ describe("resolveWorker", () => {
 
 describe("resolveDsModel", () => {
   test("defaults to flash", () => {
-    expect(resolveDsModel()).toEqual({ alias: "flash", id: "deepseek-v4-flash" });
+    expect(resolveDsModel()).toEqual({ alias: "flash", id: "deepseek-flash" });
+    expect(resolveDsModel("  ")).toEqual({ alias: "flash", id: "deepseek-flash" });
     expect(resolveDsModelAlias("")).toBe("flash");
   });
   test("accepts aliases and full ids", () => {
-    expect(resolveDsModel("flash").id).toBe("deepseek-v4-flash");
-    expect(resolveDsModel("pro").id).toBe("deepseek-v4-pro");
-    expect(resolveDsModel("vision").id).toBe("deepseek-v4-flash-vision-exp");
-    expect(resolveDsModel("deepseek-v4-flash-vision-exp")).toEqual({
+    expect(resolveDsModel("flash")).toEqual({ alias: "flash", id: "deepseek-flash" });
+    expect(resolveDsModel("deepseek-flash")).toEqual({ alias: "flash", id: "deepseek-flash" });
+    expect(resolveDsModel("deepseek-v4-flash")).toEqual({ alias: "flash", id: "deepseek-flash" });
+    expect(resolveDsModel("deepseek-v4.1-flash")).toEqual({ alias: "flash", id: "deepseek-flash" });
+    expect(resolveDsModel("pro")).toEqual({ alias: "pro", id: "deepseek-flash" });
+    expect(resolveDsModel("deepseek-v4-pro")).toEqual({ alias: "pro", id: "deepseek-flash" });
+    expect(resolveDsModel("deepseek-v4-pro[1m]")).toEqual({ alias: "pro", id: "deepseek-flash" });
+    expect(resolveDsModel("vision")).toEqual({ alias: "vision", id: "deepseek-flash" });
+    expect(resolveDsModel("flash-vision")).toEqual({ alias: "vision", id: "deepseek-flash" });
+    expect(resolveDsModel("deepseek-v4-flash-vision")).toEqual({
       alias: "vision",
-      id: "deepseek-v4-flash-vision-exp",
+      id: "deepseek-flash",
     });
-    expect(resolveDsModel("deepseek-v4-pro[1m]")).toEqual({
-      alias: "pro",
-      id: "deepseek-v4-pro[1m]",
+    expect(resolveDsModel("deepseek-v4-flash-vision-exp")).toEqual({
+      alias: "vision",
+      id: "deepseek-flash",
     });
   });
   test("rejects unknown", () => {
     expect(() => resolveDsModel("opus")).toThrow(/flash\|pro\|vision/);
+    expect(() => resolveDsModel("deepseek-chat")).toThrow(/flash\|pro\|vision/);
   });
 });
 
@@ -93,7 +101,7 @@ describe("promptLooksLikeVision", () => {
 });
 
 describe("claude-ds -Model", () => {
-  test("shim path passes -Model deepseek-v4-flash by default", () => {
+  test("shim path passes -Model deepseek-flash by default", () => {
     const prev = process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
     const prevDs = process.env.CURSOR_ROUTE_DS_MODEL;
     const prevAm = process.env.ANTHROPIC_MODEL;
@@ -107,7 +115,9 @@ describe("claude-ds -Model", () => {
         alwaysApprove: true,
       });
       expect(plan.command).toContain("-Model");
-      expect(plan.command).toContain("deepseek-v4-flash");
+      expect(plan.command).toContain("deepseek-flash");
+      expect(plan.command).not.toContain("deepseek-v4-flash");
+      expect(plan.command).not.toContain("deepseek-v4-pro");
     } finally {
       if (prev === undefined) delete process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
       else process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = prev;
@@ -117,7 +127,7 @@ describe("claude-ds -Model", () => {
       else process.env.ANTHROPIC_MODEL = prevAm;
     }
   });
-  test("vision alias maps to deepseek-v4-flash-vision-exp", () => {
+  test("vision alias maps to deepseek-flash", () => {
     const prev = process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
     process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = "/tmp/fake-claude-ds";
     try {
@@ -127,14 +137,14 @@ describe("claude-ds -Model", () => {
         alwaysApprove: true,
         model: "vision",
       });
-      expect(plan.command).toContain("deepseek-v4-flash-vision-exp");
-      expect(plan.command).not.toContain("deepseek-v4-flash'");
+      expect(plan.command).toContain("deepseek-flash");
+      expect(plan.command).not.toContain("deepseek-v4-flash-vision-exp");
     } finally {
       if (prev === undefined) delete process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
       else process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = prev;
     }
   });
-  test("pro alias maps to deepseek-v4-pro", () => {
+  test("pro alias maps to deepseek-flash", () => {
     const prev = process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
     process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = "/tmp/fake-claude-ds";
     try {
@@ -144,14 +154,14 @@ describe("claude-ds -Model", () => {
         alwaysApprove: true,
         model: "pro",
       });
-      expect(plan.command).toContain("deepseek-v4-pro");
-      expect(plan.command).not.toContain("deepseek-v4-flash");
+      expect(plan.command).toContain("deepseek-flash");
+      expect(plan.command).not.toContain("deepseek-v4-pro");
     } finally {
       if (prev === undefined) delete process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
       else process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = prev;
     }
   });
-  test("preserves deepseek-v4-pro[1m] via modelId", () => {
+  test("legacy deepseek-v4-pro[1m] modelId maps to deepseek-flash", () => {
     const prev = process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
     process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = "/tmp/fake-claude-ds";
     try {
@@ -162,7 +172,8 @@ describe("claude-ds -Model", () => {
         model: "pro",
         modelId: "deepseek-v4-pro[1m]",
       });
-      expect(plan.command).toContain("deepseek-v4-pro[1m]");
+      expect(plan.command).toContain("deepseek-flash");
+      expect(plan.command).not.toContain("deepseek-v4-pro");
     } finally {
       if (prev === undefined) delete process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
       else process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = prev;
@@ -232,8 +243,8 @@ describe("startJob product path", () => {
       expect(result.ok).toBe(true);
       if (!result.ok) return;
       expect(result.job.model).toBe("pro");
-      expect(result.command).toContain("deepseek-v4-pro");
-      expect(result.command).not.toContain("deepseek-v4-flash");
+      expect(result.command).toContain("deepseek-flash");
+      expect(result.command).not.toContain("deepseek-v4-pro");
     } finally {
       if (prev.bin === undefined) delete process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
       else process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = prev.bin;
@@ -306,7 +317,8 @@ describe("startJob product path", () => {
       expect(result.ok).toBe(true);
       if (!result.ok) return;
       expect(result.job.model).toBe("vision");
-      expect(result.command).toContain("deepseek-v4-flash-vision-exp");
+      expect(result.command).toContain("deepseek-flash");
+      expect(result.command).not.toContain("deepseek-v4-flash-vision-exp");
     } finally {
       if (prev.bin === undefined) delete process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
       else process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = prev.bin;
@@ -342,7 +354,7 @@ describe("startJob product path", () => {
       expect(result.ok).toBe(true);
       if (!result.ok) return;
       expect(result.job.model).toBe("flash");
-      expect(result.command).toContain("deepseek-v4-flash");
+      expect(result.command).toContain("deepseek-flash");
       expect(result.command).not.toContain("deepseek-v4-flash-vision-exp");
     } finally {
       if (prev.bin === undefined) delete process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
@@ -380,6 +392,45 @@ describe("startJob product path", () => {
       expect(result.ok).toBe(true);
       if (!result.ok) return;
       expect(result.job.model).toBe("flash");
+      expect(result.command).toContain("deepseek-flash");
+      expect(result.command).not.toContain("deepseek-v4-flash-vision-exp");
+    } finally {
+      if (prev.bin === undefined) delete process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
+      else process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = prev.bin;
+      if (prev.ds === undefined) delete process.env.CURSOR_ROUTE_DS_MODEL;
+      else process.env.CURSOR_ROUTE_DS_MODEL = prev.ds;
+      if (prev.am === undefined) delete process.env.ANTHROPIC_MODEL;
+      else process.env.ANTHROPIC_MODEL = prev.am;
+      if (prev.jobs === undefined) delete process.env.CURSOR_ROUTE_JOBS_DIR;
+      else process.env.CURSOR_ROUTE_JOBS_DIR = prev.jobs;
+      rmSync(jobsDir, { recursive: true, force: true });
+    }
+  });
+
+  test("--model vision records alias vision and launches deepseek-flash", () => {
+    const prev = {
+      bin: process.env.CURSOR_ROUTE_CLAUDE_DS_BIN,
+      ds: process.env.CURSOR_ROUTE_DS_MODEL,
+      am: process.env.ANTHROPIC_MODEL,
+      jobs: process.env.CURSOR_ROUTE_JOBS_DIR,
+    };
+    const jobsDir = join(tmpdir(), `cr-jobs-vision-flag-${process.pid}`);
+    mkdirSync(jobsDir, { recursive: true });
+    process.env.CURSOR_ROUTE_CLAUDE_DS_BIN = "/tmp/fake-claude-ds";
+    delete process.env.CURSOR_ROUTE_DS_MODEL;
+    delete process.env.ANTHROPIC_MODEL;
+    process.env.CURSOR_ROUTE_JOBS_DIR = jobsDir;
+    try {
+      const result = startJob({
+        prompt: "ping",
+        lane: "mid",
+        model: "vision",
+        dryRun: true,
+      });
+      expect(result.ok).toBe(true);
+      if (!result.ok) return;
+      expect(result.job.model).toBe("vision");
+      expect(result.command).toContain("deepseek-flash");
       expect(result.command).not.toContain("deepseek-v4-flash-vision-exp");
     } finally {
       if (prev.bin === undefined) delete process.env.CURSOR_ROUTE_CLAUDE_DS_BIN;
@@ -605,15 +656,16 @@ describe("deepseek adapter (dsh)", () => {
         const patch = readFileSync(join(dir, "job.dsh-patch.yml"), "utf8");
         expect(patch).toContain("agent-default-model");
         expect(patch).toContain("provider: deepseek-official");
-        expect(patch).toContain("deepseek-v4-flash");
+        expect(patch).toContain("deepseek-flash");
         expect(patch).not.toContain("deepseek-v4-pro");
+        expect(patch).not.toContain("deepseek-v4-flash-vision-exp");
         expect(patch).not.toContain("test-key");
       },
     );
     rmSync(dir, { recursive: true, force: true });
   });
 
-  test("buildLaunch: vision model → deepseek-v4-flash-vision-exp in patch", () => {
+  test("buildLaunch: vision model → deepseek-flash in patch", () => {
     const { dir, bin } = makeFakeDsh("vision");
     const promptFile = join(dir, "job.prompt");
     writeFileSync(promptFile, "ping");
@@ -632,13 +684,15 @@ describe("deepseek adapter (dsh)", () => {
           model: "vision",
         });
         const patch = readFileSync(join(dir, "job.dsh-patch.yml"), "utf8");
-        expect(patch).toContain("deepseek-v4-flash-vision-exp");
+        expect(patch).toContain("deepseek-flash");
+        expect(patch).not.toContain("deepseek-v4-flash-vision-exp");
+        expect(patch).not.toContain("deepseek-v4-pro");
       },
     );
     rmSync(dir, { recursive: true, force: true });
   });
 
-  test("buildLaunch: pro model → deepseek-v4-pro in patch", () => {
+  test("buildLaunch: pro model → deepseek-flash in patch", () => {
     const { dir, bin } = makeFakeDsh("pro");
     const promptFile = join(dir, "job.prompt");
     writeFileSync(promptFile, "ping");
@@ -657,14 +711,15 @@ describe("deepseek adapter (dsh)", () => {
           model: "pro",
         });
         const patch = readFileSync(join(dir, "job.dsh-patch.yml"), "utf8");
-        expect(patch).toContain("deepseek-v4-pro");
-        expect(patch).not.toContain("deepseek-v4-flash");
+        expect(patch).toContain("deepseek-flash");
+        expect(patch).not.toContain("deepseek-v4-pro");
+        expect(patch).not.toContain("deepseek-v4-flash-vision-exp");
       },
     );
     rmSync(dir, { recursive: true, force: true });
   });
 
-  test("buildLaunch: preserves deepseek-v4-pro[1m] in patch via modelId", () => {
+  test("buildLaunch: legacy deepseek-v4-pro[1m] modelId maps to deepseek-flash in patch", () => {
     const { dir, bin } = makeFakeDsh("pro1m");
     const promptFile = join(dir, "job.prompt");
     writeFileSync(promptFile, "ping");
@@ -684,7 +739,8 @@ describe("deepseek adapter (dsh)", () => {
           modelId: "deepseek-v4-pro[1m]",
         });
         const patch = readFileSync(join(dir, "job.dsh-patch.yml"), "utf8");
-        expect(patch).toContain("deepseek-v4-pro[1m]");
+        expect(patch).toContain("deepseek-flash");
+        expect(patch).not.toContain("deepseek-v4-pro");
       },
     );
     rmSync(dir, { recursive: true, force: true });
@@ -879,7 +935,7 @@ describe("health", () => {
   test("returns structured report", () => {
     const r = runHealth();
     expect(r.product).toBe("cursor-route");
-    expect(r.version).toBe("0.1.13");
+    expect(r.version).toBe("0.1.15");
     expect(r.checks.length).toBeGreaterThan(3);
     expect(r.checks.some((c) => c.name === "tmux")).toBe(true);
     expect(r.checks.some((c) => c.name === "cursor_cli")).toBe(true);
@@ -891,8 +947,8 @@ describe("health", () => {
     expect(r.checks.some((c) => c.name === "worker:deepseek")).toBe(true);
   });
 
-  test("config version is 0.1.13", () => {
-    expect(config.version).toBe("0.1.13");
+  test("config version is 0.1.15", () => {
+    expect(config.version).toBe("0.1.15");
   });
 
   test("OR-gate: ok can be true while worker:opencode is false", () => {

package/src/cli.ts

@@ -57,7 +57,7 @@ Usage:
 Start options:
   --worker <grok|claude-ds|openrouter|deepseek|opencode>  Worker adapter (default: grok; deepseek/opencode = opt-in)
   --lane <easy|mid|hard>                Lane → worker (easy=openrouter, mid=claude-ds, hard=grok)
-  --model <flash|pro|vision|free|provider/model>  claude-ds/deepseek: flash|pro|vision. opencode/openrouter: free (live catalog pick) or provider/model
+  --model <flash|pro|vision|free|provider/model>  claude-ds/deepseek: flash|pro|vision (aliases map to deepseek-flash). opencode/openrouter: free (live catalog pick) or provider/model
   --dir <path>                          Working directory (default: cwd)
   --ask                                 Disable always-approve for this job
   --dry-run                             Print launch command; do not start
@@ -70,7 +70,7 @@ Env:
   CURSOR_ROUTE_MAX_JOBS              Max active jobs (default: 50)
   CURSOR_ROUTE_RELAXED=1             health OK without tmux/workers (CI / infra smoke)
   CURSOR_ROUTE_ALLOW_ANTHROPIC=1     Allow mid-lane on Anthropic Claude (expensive; not default; --model ignored)
-  CURSOR_ROUTE_DS_MODEL              Default mid model flash|pro|vision (or full ids / deepseek-v4-pro[1m]); overridden by --model
+  CURSOR_ROUTE_DS_MODEL              Default mid model flash|pro|vision (aliases map to deepseek-flash); overridden by --model
   CURSOR_ROUTE_GROK_BIN              Override the grok binary path (tests / power users)
   CURSOR_ROUTE_CLAUDE_DS_BIN         Override the claude-ds binary path (tests / power users)
   CURSOR_ROUTE_DSH_BIN               Override the dsh binary path (tests / power users)

package/src/config.ts

@@ -23,10 +23,13 @@ export const WORKERS: WorkerKind[] = ["grok", "claude-ds", "openrouter", "deepse
 export const LANES: Lane[] = ["easy", "mid", "hard"];
 export const DS_MODELS: DsModelAlias[] = ["flash", "pro", "vision"];
 
+/** DeepSeek V4.1 Flash (native vision). Matches the 2026-09-10 API id. */
+export const DS_FLASH_ID = "deepseek-flash";
+
 export const DS_MODEL_IDS: Record<DsModelAlias, string> = {
-  flash: "deepseek-v4-flash",
-  pro: "deepseek-v4-pro",
-  vision: "deepseek-v4-flash-vision-exp",
+  flash: DS_FLASH_ID,
+  pro: DS_FLASH_ID,
+  vision: DS_FLASH_ID,
 };
 
 /** Prompt looks like a screenshot / image / ui mock — vision auto-pick. */
@@ -38,26 +41,35 @@ export function promptLooksLikeVision(prompt: string): boolean {
 }
 
 /**
- * Resolve --model / env to alias + concrete model id.
- * Preserves `deepseek-v4-pro[1m]` (does not silently strip the SKU).
+ * Resolve --model / env to alias + launch id.
+ * V4.1 Flash is native vision; Pro is a legacy alias, not a rotation step.
+ * All known aliases launch `deepseek-flash`.
  * Empty → Flash.
  */
 export function resolveDsModel(raw?: string | null): DsModelChoice {
   if (!raw || !raw.trim()) {
-    return { alias: "flash", id: DS_MODEL_IDS.flash };
+    return { alias: "flash", id: DS_FLASH_ID };
   }
-  const v = raw.trim().toLowerCase();
-  if (v === "flash" || v === "deepseek-v4-flash") {
-    return { alias: "flash", id: DS_MODEL_IDS.flash };
+  // Trim, lower-case, strip a trailing `[…]` SKU (`deepseek-v4-pro[1m]` → `deepseek-v4-pro`).
+  const v = raw.trim().toLowerCase().replace(/\[[^\]]*\]\s*$/, "");
+  if (
+    v === "flash" ||
+    v === "deepseek-flash" ||
+    v === "deepseek-v4-flash" ||
+    v === "deepseek-v4.1-flash"
+  ) {
+    return { alias: "flash", id: DS_FLASH_ID };
   }
   if (v === "pro" || v === "deepseek-v4-pro") {
-    return { alias: "pro", id: DS_MODEL_IDS.pro };
-  }
-  if (v === "vision" || v === "deepseek-v4-flash-vision-exp") {
-    return { alias: "vision", id: DS_MODEL_IDS.vision };
+    return { alias: "pro", id: DS_FLASH_ID };
   }
-  if (v === "deepseek-v4-pro[1m]") {
-    return { alias: "pro", id: "deepseek-v4-pro[1m]" };
+  if (
+    v === "vision" ||
+    v === "flash-vision" ||
+    v === "deepseek-v4-flash-vision" ||
+    v === "deepseek-v4-flash-vision-exp"
+  ) {
+    return { alias: "vision", id: DS_FLASH_ID };
   }
   throw new Error(`Invalid DeepSeek model ${raw}; expected flash|pro|vision`);
 }
@@ -111,7 +123,7 @@ function maxConcurrentJobsFromEnv(): number {
  */
 export const config = {
   product: "cursor-route",
-  version: "0.1.13",
+  version: "0.1.15",
   get jobsDir(): string {
     return defaultJobsDir();
   },

package/src/or-free.test.ts

@@ -62,10 +62,13 @@ describe("openrouter free catalog", () => {
     ).toBe(false);
   });
 
-  test("boost: coder/qwen outranks generic free; no hardcoded winner id", () => {
-    expect(orFreeBoost("qwen/qwen-coder:free")).toBe(20);
-    expect(orFreeBoost("acme/generic:free")).toBe(10);
-    expect(orFreeBoost("meta-llama/llama-3.3-70b-instruct:free")).toBe(20);
+  test("boost: tier ranks Qwen/GLM/Kimi above Nemotron; generic free is default", () => {
+    expect(orFreeBoost("qwen/qwen-coder:free")).toBe(100);
+    expect(orFreeBoost("z-ai/glm-5.2:free")).toBe(95);
+    expect(orFreeBoost("moonshotai/kimi-k2:free")).toBe(95);
+    expect(orFreeBoost("acme/generic:free")).toBe(40);
+    expect(orFreeBoost("meta-llama/llama-3.3-70b-instruct:free")).toBe(70);
+    expect(orFreeBoost("nvidia/nemotron-3-550b:free")).toBe(15);
   });
 
   test("rank prefers a :free coder with larger context over a generic :free", () => {
@@ -85,10 +88,83 @@ describe("openrouter free catalog", () => {
     expect(ranked[1]?.id).toBe("qwen/small-coder:free");
   });
 
+  test("all seven tier buckets", () => {
+    expect(orFreeBoost("qwen/qwen3-coder:free")).toBe(100);
+    expect(orFreeBoost("z-ai/glm-5.2:free")).toBe(95);
+    expect(orFreeBoost("moonshotai/kimi-k2:free")).toBe(95);
+    expect(orFreeBoost("deepseek/deepseek-chat:free")).toBe(90);
+    expect(orFreeBoost("acme/hy-foo:free")).toBe(90);
+    expect(orFreeBoost("meta-llama/llama-3.3-70b-instruct:free")).toBe(70);
+    expect(orFreeBoost("google/gemma-3-12b:free")).toBe(70);
+    expect(orFreeBoost("openai/gpt-oss-120b:free")).toBe(70);
+    expect(orFreeBoost("minimax/minimax-m1:free")).toBe(70);
+    expect(orFreeBoost("nvidia/nemotron-3-550b:free")).toBe(15);
+    expect(orFreeBoost("acme/generic:free")).toBe(40);
+  });
+
+  test("equal score: higher tier wins (Qwen ctx 0 vs GLM at ctx cap)", () => {
+    const ranked = rankOrFreeModels([
+      { id: "qwen/qwen3-coder:free", context_length: 0 },
+      { id: "z-ai/glm-5.2:free", context_length: 131_072 },
+    ]);
+    expect(ranked[0]?.boost).toBe(ranked[1]?.boost);
+    expect(ranked[0]?.id).toBe("qwen/qwen3-coder:free");
+  });
+
+  test("equal score same tier: id ascending", () => {
+    const ranked = rankOrFreeModels([
+      { id: "qwen/zzz-coder:free", context_length: 8_000 },
+      { id: "qwen/aaa-coder:free", context_length: 8_000 },
+    ]);
+    expect(ranked[0]?.id).toBe("qwen/aaa-coder:free");
+    expect(ranked[1]?.id).toBe("qwen/zzz-coder:free");
+  });
+
+  test("ctx bonus caps at 131072: same-tier 131072 vs 262144 ties then id asc", () => {
+    const ranked = rankOrFreeModels([
+      { id: "qwen/zzz-coder:free", context_length: 262_144 },
+      { id: "qwen/aaa-coder:free", context_length: 131_072 },
+    ]);
+    expect(ranked[0]?.boost).toBe(ranked[1]?.boost);
+    expect(ranked[0]?.id).toBe("qwen/aaa-coder:free");
+  });
+
+  test("huge Nemotron free model loses to GLM/Qwen on tier rank", () => {
+    const ranked = rankOrFreeModels([
+      { id: "nvidia/nemotron-3-550b:free", name: "Nemotron 3 550B", context_length: 262_144 },
+      { id: "z-ai/glm-5.2:free", name: "GLM 5.2", context_length: 32_768 },
+      { id: "qwen/qwen3-coder:free", name: "Qwen3 Coder", context_length: 32_768 },
+    ]);
+    expect(ranked[0]?.id).toBe("qwen/qwen3-coder:free");
+    expect(ranked[0]?.id).not.toBe("nvidia/nemotron-3-550b:free");
+    expect(ranked[1]?.id).toBe("z-ai/glm-5.2:free");
+    expect(ranked[2]?.id).toBe("nvidia/nemotron-3-550b:free");
+    expect(ranked[0]!.boost).toBeGreaterThan(ranked[2]!.boost);
+  });
+
+  test("huge ctx alone does not beat a higher tier (ctx bonus caps at 131072)", () => {
+    const ranked = rankOrFreeModels([
+      { id: "nvidia/nemotron-3-550b:free", context_length: 1_048_576 },
+      { id: "qwen/qwen3-coder:free", context_length: 1_000 },
+    ]);
+    expect(ranked[0]?.id).toBe("qwen/qwen3-coder:free");
+  });
+
   test("pickOrFreeModel uses the ranked catalog winner", () => {
     expect(pickOrFreeModel({ catalog: CATALOG })).toBe("qwen/qwen-coder:free");
   });
 
+  test("pickOrFreeModel winner is not a huge Nemotron when GLM/Qwen are free", () => {
+    const winner = pickOrFreeModel({
+      catalog: [
+        { id: "nvidia/nemotron-3-550b:free", name: "Nemotron 3 550B", context_length: 262_144 },
+        { id: "z-ai/glm-5.2:free", name: "GLM 5.2", context_length: 32_768 },
+      ],
+    });
+    expect(winner).not.toBe("nvidia/nemotron-3-550b:free");
+    expect(winner).toBe("z-ai/glm-5.2:free");
+  });
+
   test("offline / empty catalog falls back to the OpenRouter router", () => {
     expect(pickOrFreeModel({ catalog: [] })).toBe(OPENROUTER_FALLBACK_MODEL);
     expect(OPENROUTER_FALLBACK_MODEL).toBe("openrouter/free");
@@ -175,6 +251,18 @@ describe("openRouterModel live pick", () => {
       expect(existsSync(cache)).toBe(false);
     });
   });
+
+  test("ranked pick is cached; offline second call returns the cached id", () => {
+    isolate(() => {
+      process.env.CURSOR_ROUTE_OR_CATALOG_JSON = JSON.stringify({ data: CATALOG });
+      expect(openRouterModel("free")).toBe("qwen/qwen-coder:free");
+      delete process.env.CURSOR_ROUTE_OR_REFRESH;
+      process.env.CURSOR_ROUTE_OR_OFFLINE = "1";
+      delete process.env.CURSOR_ROUTE_OR_CATALOG_JSON;
+      expect(existsSync(process.env.CURSOR_ROUTE_OR_CACHE_PATH!)).toBe(true);
+      expect(openRouterModel("free")).toBe("qwen/qwen-coder:free");
+    });
+  });
 });
 
 describe("openrouter health stays offline", () => {

package/src/or-free.ts

@@ -35,8 +35,6 @@ export interface OrFreePick {
 
 const EXCLUDE_RE =
   /lyria|whisper|tts|embed|embedding|image|vision-only|audio|diffusion|flux|stable-diffusion|moderation/i;
-const BOOST_RE =
-  /coder|instruct|chat|nemotron|qwen|llama|gemma|gpt-oss|kimi|glm|deepseek/i;
 
 function orModelsUrl(): string {
   const base = (process.env.OPENROUTER_BASE_URL || "https://openrouter.ai/api/v1")
@@ -88,9 +86,22 @@ export function isOrFreeModel(m: OrModel): boolean {
   return tagged || pricedFree;
 }
 
-/** Higher boost wins. Coding/chat families beat generic free. No hardcoded id. */
+/**
+ * Toolkit-parity tier for a live OpenRouter free model id (first match wins).
+ * Huge general free models (Nemotron) rank well below coding/chat families so
+ * the easy-lane live pick does not land on a slow 550B generalist. Order
+ * matters: `qwen` is checked before `nemotron`, and the generic coding bucket
+ * (70) sits before `nemotron` exactly like the toolkit, so
+ * `nvidia/nemotron-3-550b:free` scores 15 — never 70. No hardcoded model id.
+ */
 export function orFreeBoost(id: string): number {
-  return BOOST_RE.test(id) ? 20 : 10;
+  if (/qwen/i.test(id)) return 100;
+  if (/(^|\/)z-ai\/glm|glm/i.test(id)) return 95;
+  if (/kimi|moonshot/i.test(id)) return 95;
+  if (/deepseek|hy-/i.test(id)) return 90;
+  if (/coder|instruct|chat|llama|gemma|gpt-oss|minimax/i.test(id)) return 70;
+  if (/nemotron/i.test(id)) return 15;
+  return 40;
 }
 
 export function rankOrFreeModels(models: OrModel[]): Array<{
@@ -110,16 +121,22 @@ export function rankOrFreeModels(models: OrModel[]): Array<{
       continue;
     }
     const ctx = Number.isFinite(Number(m.context_length)) ? Number(m.context_length) : 0;
+    const tier = orFreeBoost(id);
+    // Cap the context bonus at 131072 so a huge Nemotron ctx cannot outrank a
+    // higher tier on context alone: score = tier + min(ctx, 131072)/131072 * 5.
+    const score = tier + (Math.min(ctx, 131_072) / 131_072) * 5;
     out.push({
       id,
       name: (m.name || raw).trim(),
       context_length: ctx,
-      boost: orFreeBoost(id),
+      boost: score,
     });
   }
+  // Sort like the toolkit: score desc, then tier desc, then id asc.
   out.sort((a, b) => {
     if (b.boost !== a.boost) return b.boost - a.boost;
-    if (b.context_length !== a.context_length) return b.context_length - a.context_length;
+    const tierDiff = orFreeBoost(b.id) - orFreeBoost(a.id);
+    if (tierDiff !== 0) return tierDiff;
     return a.id.localeCompare(b.id);
   });
   return out;