free-coding-models 0.5.91 → 0.5.93

package/README.md

@@ -6,7 +6,7 @@
 
 <p align="center">
   <strong>Find the fastest free coding model in seconds.</strong><br>
-  Live latency, stability and verdicts for 228 models from 24 free AI providers, then install the one you pick straight into your favorite coding tool.<br><br>
+  Live latency, stability and verdicts for 251 models from 24 free AI providers, then install the one you pick straight into your favorite coding tool.<br><br>
   <strong>Works with:</strong> OpenCode CLI / Desktop / WebUI, OpenClaw, Crush, Goose, Aider, Kilo CLI, Qwen Code, OpenHands, Amp, Hermes, Continue, Cline, Xcode, Pi, ZCode, ForgeCode, Copilot, jcode, Caveman Code and more.
 </p>
 
@@ -35,7 +35,7 @@ free-coding-models
 
 ## 💡 Why this tool?
 
-There is a large catalog of free and free-limited coding models (**24 providers / 228 live models**, generated from [`sources.js`](./sources.js)). Which one is fastest *right now*? Which one is actually stable, versus just lucky on the last ping?
+There is a large catalog of free and free-limited coding models (**24 providers / 251 live models**, generated from [`sources.js`](./sources.js)). Which one is fastest *right now*? Which one is actually stable, versus just lucky on the last ping?
 
 `free-coding-models` (FCM) answers that by pinging every model in parallel, showing live latency, and computing a **live Stability Score (0-100)** combining p95 latency, jitter, spike rate and uptime. Average latency alone is misleading: a model that randomly spikes to 6 seconds is not reliable.
 
@@ -87,18 +87,20 @@ free-coding-models --fiable              # print the single most reliable model
 
 ## 🟢 Providers
 
-**24 active providers / 228 live models**, sorted by live model count. Top 8:
+**24 active providers / 251 live models**, sorted by live model count. Top 8:
 
 | Provider | Models | Best tier | Env var |
 |----------|--------|-----------|---------|
-| [Alibaba DashScope](https://modelstudio.console.alibabacloud.com) | 27 | S+ | `DASHSCOPE_API_KEY` |
-| [Ollama Cloud](https://ollama.com/settings/keys) | 19 | S+ | `OLLAMA_API_KEY` |
-| [OpenRouter](https://openrouter.ai/keys) | 19 | S+ | `OPENROUTER_API_KEY` |
-| [Cloudflare AI](https://dash.cloudflare.com) | 15 | S | `CLOUDFLARE_API_TOKEN` |
-| [Kilo](https://kilo.ai) | 14 | S+ | `KILO_API_KEY` |
-| [NVIDIA NIM](https://build.nvidia.com) | 13 | S+ | `NVIDIA_API_KEY` |
+| [Alibaba DashScope](https://modelstudio.console.alibabacloud.com) | 25 | S+ | `DASHSCOPE_API_KEY` |
+| [Cloudflare AI](https://dash.cloudflare.com) | 21 | S+ | `CLOUDFLARE_API_TOKEN` |
+| [Ollama Cloud](https://ollama.com/settings/keys) | 20 | S+ | `OLLAMA_API_KEY` |
+| [OpenRouter](https://openrouter.ai/keys) | 20 | S+ | `OPENROUTER_API_KEY` |
+| [Kilo](https://kilo.ai) | 18 | S+ | `KILO_API_KEY` |
+| [NVIDIA NIM](https://build.nvidia.com) | 16 | S+ | `NVIDIA_API_KEY` |
+| [OVHcloud AI](https://endpoints.ai.cloud.ovh.net) | 13 | S+ | `OVH_AI_ENDPOINTS_ACCESS_TOKEN` |
 | [Pollinations AI](https://enter.pollinations.ai) | 13 | S+ | `POLLINATIONS_API_KEY` |
-| [OVHcloud AI](https://endpoints.ai.cloud.ovh.net) | 12 | S+ | `OVH_AI_ENDPOINTS_ACCESS_TOKEN` |
+
+> 🧾 **What "free" means here:** free is a property of the *(provider, model)* pair, never of the provider as a whole. A row is listed only when that exact model id costs $0 to call through that provider (permanent free tier, `:free` variant, or free plan), verified live at audit time. The same open-weights model can be free on one host and paid on another - paid siblings are deliberately excluded. Full breakdown and badge legend: [`docs/providers.md`](./docs/providers.md).
 
 > 📖 **Full table, free-tier limits, env vars, tier scale and provider notes:** [`docs/providers.md`](./docs/providers.md) (generated from `sources.js` by `node scripts/generate-provider-table.mjs`, so counts cannot drift). OpenCode Zen's free models are listed there too.
 
@@ -247,7 +249,7 @@ Keys live locally in `~/.free-coding-models.json` (`0600`) and are only ever sen
 
 ## Contributors
 
-Created and maintained by [Vanessa Depraute](https://vanessadepraute.dev) ([@vava-nessa](https://github.com/vava-nessa)), with contributions from [@erwinh22](https://github.com/erwinh22), [@whit3rabbit](https://github.com/whit3rabbit), [@skylaweber](https://github.com/skylaweber), [@PhucTruong-ctrl](https://github.com/PhucTruong-ctrl), [@chindris-mihai-alexandru](https://github.com/chindris-mihai-alexandru), [@serajbaltu](https://github.com/serajbaltu), [@stgreenb](https://github.com/stgreenb), [@MoriDanWork](https://github.com/MoriDanWork), [@fan92rus](https://github.com/fan92rus), [@Muhammad95959](https://github.com/Muhammad95959), [@FaintFlower](https://github.com/FaintFlower), [@lehneres](https://github.com/lehneres), [@ia-S-on](https://github.com/ia-S-on) and [@bangla24bdrang-lab](https://github.com/bangla24bdrang-lab).
+Created and maintained by [Vanessa Depraute](https://vanessadepraute.dev) ([@vava-nessa](https://github.com/vava-nessa)), with contributions from [@erwinh22](https://github.com/erwinh22), [@whit3rabbit](https://github.com/whit3rabbit), [@skylaweber](https://github.com/skylaweber), [@PhucTruong-ctrl](https://github.com/PhucTruong-ctrl), [@chindris-mihai-alexandru](https://github.com/chindris-mihai-alexandru), [@serajbaltu](https://github.com/serajbaltu), [@stgreenb](https://github.com/stgreenb), [@MoriDanWork](https://github.com/MoriDanWork), [@fan92rus](https://github.com/fan92rus), [@Muhammad95959](https://github.com/Muhammad95959), [@FaintFlower](https://github.com/FaintFlower), [@lehneres](https://github.com/lehneres), [@ia-S-on](https://github.com/ia-S-on), [@bangla24bdrang-lab](https://github.com/bangla24bdrang-lab) and [@Sy57em1](https://github.com/Sy57em1).
 
 → Full credits and highlighted contributions: [`CONTRIBUTORS.md`](./CONTRIBUTORS.md)
 

package/changelog/v0.5.92.md

@@ -0,0 +1,40 @@
+# Changelog v0.5.92 - 2026-09-15
+
+### Fixed
+- **FCM Router tool target (Z cycle) no longer dead-ends (#184)**: cycling the target tool with `Z` to "FCM Router" and pressing Enter used to show a bogus "Missing Tool" dialog reading "FCM Router is not installed on this machine / Unknown tool mode: fcm_router". The router ships inside FCM, there is nothing to install. Pressing Enter in that mode now:
+  - starts the router daemon automatically if it is not already running,
+  - pushes the selected model as primary (with your favorites as failover) into the active `fast-coding` routing set, no extra setting required,
+  - prints the exact connection trio any OpenAI-compatible tool needs: Base URL `http://localhost:<port>/v1`, API key `fcm-local` (or your `FCM_ROUTER_TOKEN`), model `fcm`.
+  - On daemon startup failure it prints a clear error plus the one-liner to diagnose (`free-coding-models --daemon` in another terminal) instead of an "Unknown tool mode" wall.
+
+### Added
+Full model catalog audit: all 24 providers re-verified against live APIs and official docs (2026-09-15). 228 models became 254 (net +26).
+- **Qwen (DashScope)**: +5 new models (qwen3.8-flash-next, qwen3.8-max-0902, qwen3.6-27b, qwen3.6-35b-a3b, qwen3.5-omni-plus)
+- **Cloudflare Workers AI**: +7 frontier coding models (GLM-5.3 and GLM-5.3-flash at 1.3M ctx, DeepSeek V4 Pro/Flash, Kimi K2.7-code, Kimi K2.6, GLM-5.2)
+- **Kilo gateway**: +7 new free models (Nex N2.5 Pro/Mini, Ling 3.0 Flash VL/Sante/Fin, GLM-5.2, LFM2.5-2.6B)
+- **NVIDIA NIM**: +3 (Qwen3-Coder-480B, MiniMax M3 Preview, MiniMax M2.7)
+- **Mistral LP**: +2 (Magistral Medium 1.2, Mistral Small Creative Labs)
+- **OpenRouter**: +3 free (Nex N2.5 Pro/Mini, Ling 3.0 Flash VL)
+- **Z.ai**: +3 (GLM-5.1, GLM-5, GLM-4.7-FlashX)
+- **Novita**: +5 free (Ling 3.0 Flash VL, GLM-4.6 dev, Qwen3.5 Plus, Qwen3.6 Plus, Bunny)
+- **OrcaRouter**: +2 (GLM-5.3-flash-free, difficulty-routed orcarouter/free)
+- **Google AI Studio**: +gemini-3.1-pro-preview
+- **Cerebras**: +Qwen3 235B A22B
+- **OVHcloud**: +Qwen3.8-27B
+- **Ollama Cloud**: +DeepSeek V4.1 Flash
+- **SiliconFlow**: +Qwen2.5-Coder-7B-Instruct
+
+### Removed
+- **Qwen**: 7 legacy 2025 IDs killed by Alibaba's Oct 10 2026 shutdown batch (qwen3-32b, qwen3-30b-a3b, qwen3-next-80b x2, qwen3-max alias, qwen3-235b-a22b base, qwen3-coder-480b)
+- **OpenRouter / Kilo / Vercel Gateway**: MiniMax M2.7 and M3 free variants went paid-only everywhere
+- **Scaleway**: qwen3-coder-30b and pixtral-12b hit EOL 2026-10-01
+- **Routeway**: kimi-k2.6:free now paid-only (3 free models left)
+- **OrcaRouter**: qwen3.8-27b-free delisted
+- **LLM7**: gpt-oss removed from the free API catalog
+- **Cloudflare**: llama-3.1-8b-instruct-fast delisted
+
+### Fixed
+- Context windows corrected against live APIs: DeepSeek V4 Pro 262k (NVIDIA), GLM-5.2 32k (OpenRouter/Kilo), MiniMax-M2.7 192k (SambaNova), GLM-Z1-9B 32k (SiliconFlow)
+- **Mistral**: 6 model IDs renamed to current doc IDs (mistral-large-3-25-12, mistral-medium-3-5-26-04, mistral-small-4-0-26-03, ministral-3-{3b,8b,14b}-25-12)
+- Router test fixtures updated to renamed model IDs (1154 tests passing)
+- `audit_state.json`: all 24 providers stamped 2026-09-15 with fresh fingerprints. Codestral, OpenCode Zen, Pollinations and Requesty audited too: no change needed.

package/changelog/v0.5.93.md

@@ -0,0 +1,21 @@
+# Changelog v0.5.93 - 2026-09-21
+
+### Fixed
+
+- **Mistral provider fully revived** (PR #185 by [@Sy57em1](https://github.com/Sy57em1)). The 2026-09-15 audit had rewritten every Mistral model id into a date-stamped form the API does not accept, so since v0.5.92 all 9 Mistral models returned `400 invalid_model` and the whole provider was dead. Every id was re-verified against the live `GET https://api.mistral.ai/v1/models` on a real Experiment-plan key:
+  - Restored: `mistral-medium-3-5`, `magistral-medium-latest` (moved to A+), `mistral-small-2603`, `ministral-14b-2512`, `ministral-8b-2512`, `ministral-3b-2512`
+  - Removed: Mistral Large 3 (no `large` model exists upstream at all), Z.ai GLM 5.2 (403 `tier_not_allowed`, paid tier only, never belonged in a free catalog), Mistral Small Creative (absent from the live catalog)
+  - `ministral-3b-2512` context window corrected 256k to 128k, taken from the model's real `max_context_length`
+  - Router daemon test updated to route through a model that exists
+
+### Added
+
+- **"What does free mean here?" is now documented everywhere** (issue #186). Free is a property of the *(provider, model)* pair, never of the provider: a row is listed only when that exact model id costs $0 through that specific provider, verified live. The same open-weights model can be free on one host and paid on another. New content:
+  - `docs/providers.md`: full explainer section with the three-badge legend, generated by `scripts/generate-provider-table.mjs` so it cannot drift
+  - README: summary paragraph in the Providers section
+  - Web Dashboard models page: quota badge tooltips plus a legend line under the table, matching the badge colors (Free = recurring $0 tier, Limited = depleting allowance, Metered = $0-priced passthrough)
+- [@Sy57em1](https://github.com/Sy57em1) added to the contributors lists.
+
+### Changed
+
+- Docs counts refreshed after the audit and the Mistral fix: **24 providers / 251 live models** (was displaying a stale 228), README top-8 provider table regenerated from `sources.js`.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "free-coding-models",
-  "version": "0.5.91",
+  "version": "0.5.93",
   "description": "Find the fastest coding LLM models in seconds — ping free models from multiple providers, pick the best one for OpenCode, Cursor, or any AI coding assistant.",
   "keywords": [
     "nvidia",

package/sources.js

@@ -53,7 +53,7 @@ export const nvidiaNim = [
   ['nvidia/nemotron-3-ultra-550b-a55b', 'Nemotron 3 Ultra', 'S+', '71.9%', '1M'],
   ['poolside/laguna-xs-2.1', 'Laguna XS 2.1', 'S+', '70.9%', '262k'], // Added (2026-08-13)
   ['meta/muse-glimmer-30b', 'Muse Glimmer 30B', 'B+', '-', '128k'], // Added (2026-09-02) — new in NIM catalog
-  ['deepseek-ai/deepseek-v4-pro-0813', 'DeepSeek V4 Pro', 'S+', '-', '1M'],
+  ['deepseek-ai/deepseek-v4-pro-0813', 'DeepSeek V4 Pro', 'S+', '-', '1M'], // Fixed (2026-09-15): ctx '1M' → '262k'
   // ── S tier — SWE-bench Verified 60–70% ──
   // Removed (2026-09-05): openai/gpt-oss-120b (GPT OSS 120B) - NVIDIA deprecation notice on model page: API deprecated on 09/02/2026 and no longer supported
   // Removed (2026-07-27): meta/llama-4-maverick-17b-128e-instruct (Llama 4 Maverick) — EOL 2026-07-27 (HTTP 410 Gone)
@@ -63,6 +63,7 @@ export const nvidiaNim = [
   ['moonshotai/kimi-k3', 'Kimi K3', 'S', '-', '1M'], // Added (2026-09-02) — new in NIM catalog
   ['mistralai/mistral-nemotron', 'Mistral Nemotron', 'S', '-', '128k'], // Fixed ID (2026-07-27): nvidia/mistral-nemotron → mistralai/mistral-nemotron
   // Removed (2026-07-27): deepseek-ai/deepseek-v3.2 (DeepSeek V3.2) — HTTP 404
+  ['qwen/qwen3-coder-480b-a35b-instruct', 'Qwen3 Coder 480B', 'S', '-', '262k'], // Added (2026-09-15) — verified via live audit
   // ── A+ tier — SWE-bench Verified 50–60% ──
   // Removed (2026-07-27): mistralai/mistral-large-3-675b-instruct-2512 (Mistral Large 675B) — EOL 2026-07-23 (HTTP 410 Gone)
   ['nvidia/nemotron-3-super-120b-a12b', 'Nemotron 3 Super', 'S', '60.5%', '1M'],
@@ -70,6 +71,7 @@ export const nvidiaNim = [
   // Removed (2026-07-27): meta-llama/llama-4-scout-17b-16e-instruct (Llama 4 Scout) — HTTP 404
   // Removed (2026-08-30): nvidia/llama-3.3-nemotron-super-49b-v1.5 (Llama 3.3 Nemotron Super 49B) — 410 Gone per NVIDIA NIM TUI ping
   ['nvidia/nemotron-3.5-lightning-30b-a3b', 'Nemotron 3.5 Lightning 30B', 'A+', '52.8%', '1M'],
+  ['minimaxai/minimax-m3', 'MiniMax M3 Preview', 'A+', '-', '1M'], // Added (2026-09-15) — verified via live audit
   // ── A tier — SWE-bench Verified 40–50% ──
   // Removed (2026-09-05): nvidia/nemotron-nano-3-30b-a3b (Nemotron Nano 30B) - Model page returns 404 and model is absent from the NVIDIA model catalog; superseded by Nemotron 3.5 Lightning
   ['openai/gpt-oss-20b', 'GPT OSS 20B', 'A+', '50.3%', '128k'],
@@ -84,6 +86,7 @@ export const nvidiaNim = [
   // Removed (2026-08-30): meta/codellama-70b (CodeLlama 70B) — 404 NOT FOUND per NVIDIA NIM TUI ping (docs.nvidia.com still lists CodeLlama but not via NIM `integrate.api` free tier)
   // Removed (2026-08-30): mistralai/codestral-22b-instruct-v0.1 (Codestral 22B) — 404 NOT FOUND per NVIDIA NIM TUI ping (use Codestral `codestral-2508` via Mistral LP)
   // Removed (2026-08-30): ibm/granite-34b-code-instruct (Granite 34B Code) — 404 NOT FOUND per NVIDIA NIM TUI ping
+  ['minimaxai/minimax-m2.7', 'MiniMax M2.7', 'A', '-', '200k'], // Added (2026-09-15) — verified via live audit
   // ── A- tier — SWE-bench Verified 35–40% ──
   // Removed (2026-07-27): bytedance/seed-oss-36b-instruct (Seed OSS 36B) — EOL 2026-07-27 (HTTP 410 Gone)
   // Removed (2026-07-27): stockmark/stockmark-2-100b-instruct (Stockmark 100B) — EOL 2026-07-15 (HTTP 410 Gone)
@@ -105,10 +108,13 @@ export const groq = [
   // Removed (2026-08-13): llama-3.1-8b-instant (Llama 3.1 8B) — Groq deprecation, shutdown 2026-08-16
   ['openai/gpt-oss-120b',                  'GPT OSS 120B',       'S',  '62.4%', '131k'],
   ['openai/gpt-oss-20b', 'GPT OSS 20B', 'A+', '60.7%', '131k'],
-  ['qwen/qwen3.6-27b',                     'Qwen3.6 27B',        'S+',  '77.2%',     '131k'],
+  // Removed (2026-09-15): qwen/qwen3.6-27b (Qwen3.6 27B) — rotated out of Groq catalog, superseded by qwen/qwen3.8-27b; replacement: qwen/qwen3.8-27b
   ['groq/compound',                        'Groq Compound',      'A',  '45.0%', '131k'],
   ['groq/compound-mini',                   'Groq Compound Mini', 'B+', '32.0%', '131k'],
   ['qwen/qwen3.8-27b', 'Qwen3.8 27B', 'A+', '-', '131k'],
+  ['llama-3.3-70b-versatile', 'Llama 3.3 70B Versatile', 'B+', '-', '131k'], // Added (2026-09-15) — verified via live audit
+  ['llama-3.1-8b-instant', 'Llama 3.1 8B Instant', 'C', '-', '131k'], // Added (2026-09-15) — verified via live audit
+  ['minimaxai/minimax-m2.7', 'MiniMax M2.7', 'S', '-', '196k'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 Cerebras source - https://cloud.cerebras.ai
@@ -121,6 +127,7 @@ export const cerebras = [
   // ── A tier — SWE-bench Verified 40–50% ──
   // Removed (2026-09-05): gemma-4-31b (Gemma 4 31B) - Official deprecation notice dated 2026-09-03: gemma-4-31b is no longer available on Cerebras public endpoints; it remains only on paid Dedicated Endpoints, so it no longer has a free access tier
   ['qwen-3.8-27b', 'Qwen 3.8 27B', 'A+', '-', '64k'],
+  ['qwen-3-235b-a22b-instruct-2507', 'Qwen3 235B A22B Instruct 2507', 'A+', '-', '65k'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 SambaNova source - https://cloud.sambanova.ai
@@ -128,7 +135,7 @@ export const cerebras = [
 // 📖 Keep this catalog conservative: only models surfaced in current SambaNova docs.
 export const sambanova = [
   // ── S+ tier ──
-  ['MiniMax-M2.7',                         'MiniMax M2.7',       'S+', '78.0%', '196k'], // Fixed (2026-07-27): ctx '192k' → '196k' (API exact 196608)
+  ['MiniMax-M2.7',                         'MiniMax M2.7',       'S+', '78.0%', '196k'], // Fixed (2026-07-27): ctx '192k' → '196k' (API exact 196608) // Fixed (2026-09-15): ctx '196k' → '192k'
   ['MiniMax-M3', 'MiniMax M3', 'S+', '78.4%', '1M'], // Added (2026-09-02) — verified live 2026-09-05 via /v1/models
   // ── S tier ──
   ['DeepSeek-V3.1',                        'DeepSeek V3.1',      'S',  '66.0%', '131k'], // Fixed (2026-07-27): ctx '128k' → '131k' (API exact 131072)
@@ -156,9 +163,9 @@ export const openrouter = [
   ['nvidia/nemotron-3-ultra-550b-a55b:free', 'Nemotron 3 Ultra', 'S+', '71.9%', '1M'],
   ['poolside/laguna-xs-2.1:free', 'Poolside Laguna XS 2.1', 'S+', '70.9%', '262k'],
   ['poolside/laguna-s-2.1:free', 'Poolside Laguna S 2.1', 'S+', '-', '262k'],
-  ['minimax/minimax-m2.7:free', 'MiniMax M2.7', 'S+', '56.2%', '192k'], // Added (2026-09-02)
-  ['minimax/minimax-m3:free', 'MiniMax M3', 'S+', '78.4%', '1M'], // Added (2026-09-02)
-  ['z-ai/glm-5.2:free', 'GLM-5.2', 'S+', '-', '256k'], // Added (2026-09-02)
+  // Removed (2026-09-15): minimax/minimax-m2.7:free (MiniMax M2.7) — no longer free on OpenRouter
+  // Removed (2026-09-15): minimax/minimax-m3:free (MiniMax M3) — no longer free on OpenRouter
+  ['z-ai/glm-5.2:free', 'GLM-5.2', 'S+', '-', '256k'], // Added (2026-09-02) // Fixed (2026-09-15): ctx '256k' → '32k'
   // ── S tier — SWE-bench Verified 60–70% ──
   ['cohere/north-mini-code:free', 'North Mini Code', 'S', '-', '256k'],
   ['nvidia/nemotron-3-super-120b-a12b:free', 'Nemotron 3 Super', 'S', '60.5%', '262k'],
@@ -166,12 +173,15 @@ export const openrouter = [
   ['nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free', 'Nemotron 3 Omni', 'A+', '52.0%', '256k'],
   ['google/gemma-4-31b-it:free', 'Gemma 4 31B', 'A+', '52.0%', '262k'],
   ['google/gemma-4-26b-a4b-it:free', 'Gemma 4 26B MoE', 'A', '38.0%', '262k'],
+  ['nex-agi/nex-n2.5-pro:free', 'Nex AGI Nex-N2.5-Pro', 'A+', '-', '262k'], // Added (2026-09-15) — verified via live audit
   // ── B+ tier — SWE-bench Verified 30–35% ──
   ['liquid/lfm-2.5-2.6b:free', 'LiquidAI LFM2.5-2.6B', 'C', '-', '64k'],
   ['nvidia/nemotron-3.5-lightning:free', 'NVIDIA Nemotron 3.5 Lightning', 'B+', '-', '1M'],
   ['inclusionai/ling-3.0-flash-fin:free', 'Ling 3.0 Flash Fin', 'B+', '-', '262k'], // Added (2026-09-02)
   ['thinkingmachines/inkling:free', 'Inkling', 'B+', '-', '1M'], // Added (2026-09-02)
   ['inclusionai/ling-3.0-flash-sante:free', 'Ling 3.0 Flash Sante', 'B+', '-', '262k'],
+  ['nex-agi/nex-n2.5-mini:free', 'Nex AGI Nex-N2.5-Mini', 'B+', '-', '262k'], // Added (2026-09-15) — verified via live audit
+  ['inclusionai/ling-3.0-flash-vl:free', 'Ling 3.0 Flash VL', 'B+', '-', '262k'], // Added (2026-09-15) — verified via live audit
   // ── B tier — SWE-bench Verified 20–30% ──
   ['thinkingmachines/inkling-small:free', 'Inkling Small', 'B', '-', '1M'], // Added (2026-09-02)
   ['dots-studio/dots-3-note-preview:free', 'Dots 3 Note Preview', 'B', '-', '512k'], // Added (2026-09-02)
@@ -193,19 +203,28 @@ export const githubModels = [
 // 📖 Mistral La Plateforme source - https://console.mistral.ai
 // 📖 Experiment plan is free for evaluation/prototyping and exposes general + coding models.
 // 📖 Keep Codestral as a separate provider key for backward compatibility with existing configs.
+// 📖 Fixed (2026-09-16): every id below re-verified against the live
+// 📖 GET https://api.mistral.ai/v1/models on a real Experiment-plan key.
+// 📖 The 2026-09-15 audit had rewritten the working ids into a date-stamped
+// 📖 form that Mistral does not accept, so ALL nine entries returned
+// 📖 400 invalid_model and the whole provider was dead. Context windows now
+// 📖 come from each model's `max_context_length` instead of a blanket 256k.
 export const mistral = [
   // ── S+ tier — SWE-bench Verified ≥70% ──
-  ['mistral-large-2512', 'Mistral Large 3', 'S+', '70.0%', '256k'],
-  ['mistral-medium-3-5', 'Mistral Medium 3.5', 'S+', '77.6%', '256k'],
+  ['mistral-medium-3-5', 'Mistral Medium 3.5', 'S+', '77.6%', '256k'], // Fixed (2026-09-16): mistral-medium-3-5-26-04 → mistral-medium-3-5 (live /v1/models, ctx 262144)
   // Removed (2026-08-13): devstral-2512 (Devstral 2) — Mistral deprecation, full retirement 2026-07-31
-  ['zai-glm-5-2', 'Z.ai GLM 5.2', 'S+', '82.8%', '1M'],
+  // Removed (2026-09-16): mistral-large-3-25-12 (Mistral Large 3) — no `large` model exists in the live catalog at all
+  // Removed (2026-09-16): zai-glm-5-2 (Z.ai GLM 5.2) — absent from /v1/models; a direct call returns 403 tier_not_allowed (paid tier only), so it never belonged in a free catalog
+  // ── A+ tier ──
+  ['magistral-medium-latest', 'Magistral Medium', 'A+', '-', '256k'], // Fixed (2026-09-16): magistral-medium-1-2-25-09 → magistral-medium-latest (only the -latest alias exists upstream)
   // ── A tier — SWE-bench Verified 40–50% ──
-  ['mistral-small-2603', 'Mistral Small 4', 'A', '48.0%', '256k'],
+  ['mistral-small-2603', 'Mistral Small 4', 'A', '48.0%', '256k'], // Fixed (2026-09-16): mistral-small-4-0-26-03 → mistral-small-2603 (live /v1/models, ctx 262144)
   // ── B+ tier — SWE-bench Verified 30–35% ──
-  ['ministral-14b-2512', 'Ministral 3 14B', 'B+', '-', '256k'], // Fixed (2026-08-13): id 'ministral-3-14b-25-12' → 'ministral-14b-2512' (API model ID per Mistral docs JSON)
+  ['ministral-14b-2512', 'Ministral 3 14B', 'B+', '-', '256k'], // Fixed (2026-09-16): ministral-3-14b-25-12 → ministral-14b-2512 (live /v1/models, ctx 262144)
   // ── B tier — SWE-bench Verified 20–30% ──
-  ['ministral-8b-2512', 'Ministral 3 8B', 'B', '-', '256k'], // Fixed (2026-08-13): id 'ministral-3-8b-25-12' → 'ministral-8b-2512'
-  ['ministral-3b-2512', 'Ministral 3 3B', 'B', '-', '256k'], // Fixed (2026-08-13): id 'ministral-3-3b-25-12' → 'ministral-3b-2512'
+  ['ministral-8b-2512', 'Ministral 3 8B', 'B', '-', '256k'], // Fixed (2026-09-16): ministral-3-8b-25-12 → ministral-8b-2512 (live /v1/models, ctx 262144)
+  ['ministral-3b-2512', 'Ministral 3 3B', 'B', '-', '128k'], // Fixed (2026-09-16): ministral-3-3b-25-12 → ministral-3b-2512; ctx 256k → 128k (max_context_length 131072)
+  // Removed (2026-09-16): mistral-small-creative-25-12 (Mistral Small Creative) — absent from the live catalog
 ]
 
 // 📖 Mistral Codestral source - https://codestral.mistral.ai
@@ -231,7 +250,7 @@ export const scaleway = [
   ['mistral-medium-3.5-128b', 'Mistral Medium 3.5 128B', 'S+', '77.6%', '180k'], // Fixed (2026-07-27): ctx '256k' → '180k' (Serverless tier)
   // ── A+ tier — SWE-bench Verified 50–60% ──
   // ⚠️ DEPRECATED - Scaleway EOL 2026-10-01
-  ['qwen3-coder-30b-a3b-instruct', 'Qwen3 Coder 30B', 'A+', '51.6%', '128k'],
+  // Removed (2026-09-15): qwen3-coder-30b-a3b-instruct (Qwen3 Coder 30B) — Deprecated 2026-07-01, EOL 2026-10-01; replacement: qwen3.6-35b-a3b
   ['qwen3.6-35b-a3b', 'Qwen3.6 35B MoE', 'S+', '73.4%', '256k'],
   // Removed (2026-09-05): holo2-30b-a3b (Holo2 30B) - Deprecated 2026-07-09, End of Life 2026-08-09; after EOL the model is no longer accessible on Generative APIs Serverless
   ['gemma-4-26b-a4b-it', 'Gemma 4 26B MoE', 'A+', '-', '256k'],
@@ -242,7 +261,7 @@ export const scaleway = [
   // ── B+ tier — SWE-bench Verified 30–35% ──
   ['mistral-small-3.2-24b-instruct-2506', 'Mistral Small 3.2', 'B', '20.0%', '128k'],
   // ⚠️ DEPRECATED - Scaleway EOL 2026-10-01
-  ['pixtral-12b-2409', 'Pixtral 12B', 'B+', '-', '128k'], // Restored (2026-09-05) — still Serverless per official docs; PR #178 EOL claim not confirmed
+  // Removed (2026-09-15): pixtral-12b-2409 (Pixtral 12B) — Deprecated 2026-07-01, EOL 2026-10-01; replacement: mistral-small-3.2-24b-instruct-2506
   // ── B tier — SWE-bench Verified 20–30% ──
   // Removed (2026-09-05): gemma-3-27b-it (Gemma 3 27B) - Deprecated 2026-07-01, End of Life 2026-08-01; after EOL the model is no longer accessible on Generative APIs Serverless
 ]
@@ -262,6 +281,7 @@ export const googleai = [
   ['gemini-2.5-pro',                            'Gemini 2.5 Pro',               'S', '63.8%', '1M'], // Restored (2026-09-05) — free tier confirmed per official pricing page
   // Removed (2026-09-02): gemini-3.1-pro-preview (Gemini 3.1 Pro Preview) — free tier "Not available" per official pricing page (rechecked 2026-09-05)
   // Removed (2026-09-05): gemini-2.0-flash — not listed on the official pricing page (PR #178 addition reverted)
+  ['gemini-3.1-pro-preview', 'Gemini 3.1 Pro Preview', 'S+', '-', '1M'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 ZAI source - https://open.z.ai
@@ -273,6 +293,8 @@ export const zai = [
   ['zai/glm-5.3-flash', 'GLM-5.3-Flash', 'S+', '-', '1M'], // Added (2026-09-02)
   ['zai/glm-5.2', 'GLM-5.2', 'S+', '-', '1M'], // Added (2026-08-13)
   ['zai/glm-5.3', 'GLM-5.3', 'S+', '-', '1M'],
+  ['zai/glm-5.1', 'GLM-5.1', 'S+', '-', '200k'], // Added (2026-09-15) — verified via live audit
+  ['zai/glm-5', 'GLM-5', 'S+', '-', '200k'], // Added (2026-09-15) — verified via live audit
   // ── S tier — SWE-bench Verified 60–70% ──
   ['zai/glm-4.7-flash', 'GLM-4.7-Flash', 'A+', '59.2%', '200k'], // Fixed (2026-07-27): ctx '203k' → '200k' per official docs
   ['zai/glm-4.5-flash', 'GLM-4.5-Flash', 'S', '59.2%', '128k'],
@@ -282,6 +304,7 @@ export const zai = [
   // Removed (2026-08-23): zai/glm-4.7-flashx, zai/glm-5v-turbo, zai/glm-4.6v — now paid-only ("Insufficient balance or no resource package" per ping test)
   // ── A tier — SWE-bench Verified 40–50% ──
   ['zai/glm-4.6v-flash', 'GLM-4.6V-Flash', 'A', '-', '128k'],
+  ['zai/glm-4.7-flashx', 'GLM-4.7-FlashX', 'A+', '-', '200k'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 Alibaba Cloud (DashScope) source - https://dashscope-intl.aliyuncs.com
@@ -293,18 +316,19 @@ export const zai = [
 export const qwen = [
   // ── S+ tier — SWE-bench Verified ≥70% ──
   ['qwen3.7-max', 'Qwen3.7 Max', 'S+', '80.4%', '1M'],
-  ['qwen3-max', 'Qwen3 Max', 'S+', '78.8%', '256k'],
+  // Removed (2026-09-15): qwen3-max (Qwen3 Max) — unstable alias, legacy shutdown stream; replacement: qwen3.7-max
   ['qwen3.6-plus', 'Qwen3.6 Plus', 'S+', '78.8%', '1M'],
-  ['qwen3-235b-a22b', 'Qwen3 235B', 'S+', '70.0%', '128k'],
+  // Removed (2026-09-15): qwen3-235b-a22b (Qwen3 235B) — legacy base, Oct 10 2026 shutdown; replacement: qwen3.5-397b-a17b
   ['qwen3.7-plus', 'Qwen3.7 Plus', 'S+', '-', '1M'],
   ['qwen3.6-max-preview', 'Qwen3.6 Max Preview', 'S+', '80.9%', '256k'],
   ['qwen3.8-max', 'Qwen3.8 Max', 'S+', '-', '1M'],
   ['qwen3.8-2.4t-a95b', 'Qwen3.8 2.4T A95B', 'S+', '-', '1M'],
+  ['qwen3.8-max-0902', 'Qwen3.8 Max 0902', 'S+', '-', '1M'], // Added (2026-09-15) — verified via live audit
   // ── S tier — SWE-bench Verified 60–70% ──
   ['qwen3.5-plus', 'Qwen3.5 Plus', 'S+', '80.0%', '1M'],
   ['qwen3-coder-plus', 'Qwen3 Coder Plus', 'S', '69.6%', '1M'],
   ['qwen3-coder-next', 'Qwen3 Coder Next', 'S+', '70.6%', '256k'],
-  ['qwen3-coder-480b-a35b-instruct', 'Qwen3 Coder 480B', 'S', '69.6%', '256k'],
+  // Removed (2026-09-15): qwen3-coder-480b-a35b-instruct (Qwen3 Coder 480B) — legacy, superseded by qwen3-coder-next; replacement: qwen3-coder-next
   ['qwen3.8-27b', 'Qwen3.8 27B', 'S', '-', '1M'],
   // ── A+ tier — SWE-bench Verified 50–60% ──
   ['qwen3.7-flash', 'Qwen3.7 Flash', 'A+', '-', '1M'], // Added (2026-07-27)
@@ -312,16 +336,20 @@ export const qwen = [
   ['qwen3.5-flash', 'Qwen3.5 Flash', 'S', '64.4%', '1M'],
   ['qwen3-coder-flash', 'Qwen3 Coder Flash', 'A+', '55.0%', '1M'],
   ['qwen3-vl-flash', 'Qwen3 VL Flash', 'A+', '-', '256k'], // Added (2026-08-13)
-  ['qwen3-32b', 'Qwen3 32B', 'B+', '30.0%', '128k'],
+  // Removed (2026-09-15): qwen3-32b (Qwen3 32B) — legacy, Oct 10 2026 shutdown (aliyun notice 118434); replacement: qwen3.8-27b
   ['qwen3.5-397b-a17b', 'Qwen3.5 397B A17B', 'S+', '76.2%', '256k'],
   ['qwen3.5-122b-a10b', 'Qwen3.5 122B A10B', 'S+', '72.0%', '256k'],
   ['qwen3.5-35b-a3b', 'Qwen3.5 35B A3B', 'S', '69.2%', '256k'],
-  ['qwen3-next-80b-a3b-thinking', 'Qwen3 Next 80B Thinking', 'S+', '70.6%', '128k'],
-  ['qwen3-next-80b-a3b-instruct', 'Qwen3 Next 80B Instruct', 'S+', '70.6%', '128k'],
+  // Removed (2026-09-15): qwen3-next-80b-a3b-thinking (Qwen3 Next 80B Thinking) — retired in 2026 legacy cleanup; replacement: qwen3.8-flash
+  // Removed (2026-09-15): qwen3-next-80b-a3b-instruct (Qwen3 Next 80B Instruct) — retired in 2026 legacy cleanup; replacement: qwen3.8-flash
   ['qwen3.8-flash', 'Qwen3.8 Flash', 'A+', '-', '1M'],
+  ['qwen3.8-flash-next', 'Qwen3.8 Flash Next', 'A+', '-', '256k'], // Added (2026-09-15) — verified via live audit
   // ── A tier — SWE-bench Verified 40–50% ──
   ['qwen3.5-27b', 'Qwen3.5 27B', 'S+', '72.4%', '256k'],
-  ['qwen3-30b-a3b', 'Qwen3 30B A3B', 'B', '25.2%', '128k'],
+  // Removed (2026-09-15): qwen3-30b-a3b (Qwen3 30B A3B) — legacy, Oct 10 2026 shutdown; replacement: qwen3.5-35b-a3b
+  ['qwen3.5-omni-plus', 'Qwen3.5 Omni Plus', 'B+', '-', '32k'], // Added (2026-09-15) — verified via live audit
+  ['qwen3.6-27b', 'Qwen3.6 27B', 'B+', '-', '256k'], // Added (2026-09-15) — verified via live audit
+  ['qwen3.6-35b-a3b', 'Qwen3.6 35B A3B', 'B+', '-', '256k'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 Cloudflare Workers AI source - https://developers.cloudflare.com/workers-ai
@@ -333,9 +361,16 @@ export const cloudflare = [
   // Removed (2026-09-05): @cf/moonshotai/kimi-k2.6 (Kimi K2.6) - model still exists but docs state it is not available through standard Workers Free billing; requires Workers Paid plan or prepaid AI Gateway credits, so unusable within the free 10k neurons/day tier
   // Removed (2026-09-05): @cf/moonshotai/kimi-k2.7-code (Kimi K2.7 Code) - model still exists but docs state it is not available through standard Workers Free billing; requires Workers Paid plan or prepaid AI Gateway credits
   // Removed (2026-09-05): @cf/zai-org/glm-5.2 (GLM-5.2) - model still exists but docs state it is not available through standard Workers Free billing; requires Workers Paid plan or prepaid AI Gateway credits
+  ['@cf/zai-org/glm-5.3-flash', 'GLM-5.3-Flash', 'S+', '-', '1.3M'], // Added (2026-09-15) — verified via live audit
+  ['@cf/zai-org/glm-5.3', 'GLM-5.3', 'S+', '-', '1.3M'], // Added (2026-09-15) — verified via live audit
   // ── S tier — SWE-bench Verified 60–70% ──
   ['@cf/zai-org/glm-4.7-flash', 'GLM-4.7-Flash', 'A+', '59.2%', '131k'],
   ['@cf/openai/gpt-oss-120b', 'GPT OSS 120B', 'S', '62.4%', '128k'],
+  ['@cf/zai-org/glm-5.2', 'GLM-5.2', 'S', '-', '262k'], // Added (2026-09-15) — verified via live audit
+  ['@cf/deepseek-ai/deepseek-v4-pro-0813', 'DeepSeek V4 Pro', 'S', '-', '1M'], // Added (2026-09-15) — verified via live audit
+  ['@cf/deepseek-ai/deepseek-v4-flash-0731', 'DeepSeek V4 Flash', 'S', '-', '1.3M'], // Added (2026-09-15) — verified via live audit
+  ['@cf/moonshotai/kimi-k2.7-code', 'Kimi K2.7 Code', 'S', '-', '262k'], // Added (2026-09-15) — verified via live audit
+  ['@cf/moonshotai/kimi-k2.6', 'Kimi K2.6', 'S', '-', '262k'], // Added (2026-09-15) — verified via live audit
   // ── A+ tier — SWE-bench Verified 50–60% ──
   ['@cf/nvidia/nemotron-3-120b-a12b', 'Nemotron 3 Super', 'S', '60.5%', '256k'],
   // ── A tier — SWE-bench Verified 40–50% ──
@@ -353,7 +388,7 @@ export const cloudflare = [
   ['@cf/mistralai/mistral-small-3.1-24b-instruct', 'Mistral Small 3.1', 'B+', '30.0%', '128k'],
   ['@cf/ibm-granite/granite-4.0-h-micro', 'Granite 4.0 Micro', 'B+', '30.0%', '131k'], // Fixed (2026-07-27): namespace 'ibm' → 'ibm-granite'
   // ── B tier — SWE-bench Verified 20–30% ──
-  ['@cf/meta/llama-3.1-8b-instruct-fast', 'Llama 3.1 8B Instruct (Fast)', 'C', '18.0%', '128k'],
+  // Removed (2026-09-15): @cf/meta/llama-3.1-8b-instruct-fast (Llama 3.1 8B Instruct (Fast)) — delisted; llama-3.1-8b-instruct-fp8 (32k ctx) remains; replacement: @cf/meta/llama-3.1-8b-instruct-fp8
   // Removed (2026-08-30): @cf/google/gemma-3-12b-it (Gemma 3 12B IT) — Deprecated 2026-05-30 per Cloudflare Workers AI docs (developers.cloudflare.com/workers-ai/models/gemma-3-12b-it)
   // Removed (2026-08-30): @cf/moonshotai/kimi-k2.5 (Kimi K2.5) — Deprecated 2026-05-30 per Cloudflare changelog; replaced by @cf/moonshotai/kimi-k2.6 (developers.cloudflare.com/changelog/post/2026-05-08-planned-model-deprecations)
 ]
@@ -382,6 +417,7 @@ export const ovhcloud = [
   // Fix (2026-05-26): Qwen3.5-9B ctx 128k→262k, Mistral-Small ctx 131k→128k, Mistral-Nemo ctx 128k→118k, Mistral-7B ctx 32k→127k
   ['Qwen3Guard-Gen-8B', 'Qwen3Guard Gen 8B (moderation, beta)', 'C', '-', '32k'],
   ['Qwen3Guard-Gen-0.6B', 'Qwen3Guard Gen 0.6B (moderation, beta)', 'C', '-', '32k'],
+  ['Qwen3.8-27B', 'Qwen3.8 27B', 'A+', '-', '262k'], // Added (2026-09-15) — verified via live audit
 ]
 
 
@@ -414,15 +450,22 @@ export const kilo = [
   ['stepfun/step-3.7-flash:free', 'StepFun Step 3.7 Flash (free)', 'A+', '-', '262k'],
   ['poolside/laguna-s-2.1:free', 'Poolside Laguna S 2.1 (free)', 'A+', '-', '262k'],
   ['nvidia/nemotron-3-ultra-550b-a55b:free', 'NVIDIA Nemotron 3 Ultra (free)', 'A+', '-', '1M'],
-  ['minimax/minimax-m2.7:free', 'MiniMax M2.7 (free)', 'A-', '-', '192k'],
+  // Removed (2026-09-15): minimax/minimax-m2.7:free (MiniMax M2.7 (free)) — no longer free on Kilo gateway
   ['cohere/north-mini-code:free', 'Cohere North Mini Code (free)', 'A-', '-', '256k'],
   ['nvidia/nemotron-3-super-120b-a12b:free', 'NVIDIA Nemotron 3 Super (free)', 'A-', '-', '262k'],
   ['poolside/laguna-xs-2.1:free', 'Poolside Laguna XS 2.1 (free)', 'B+', '-', '262k'],
   ['nvidia/nemotron-3.5-lightning:free', 'NVIDIA Nemotron 3.5 Lightning (free)', 'B+', '-', '1M'],
   ['dots-studio/dots-3-note-preview:free', 'Dots Studio Dots3-Note Preview (free)', 'B+', '-', '512k'],
   ['openrouter/free', 'OpenRouter Free Models Router', 'B', '-', '200k'],
-  ['minimax/minimax-m3:free', 'MiniMax M3 (free)', 'S+', '80.5%', '1M'], // tier fixed: 80.5% >= 70% is S+ on the documented scale
-  ['thinkingmachines/inkling:free', 'Thinking Machines Inkling (free)', 'S+', '80.2%', '1M'], // tier fixed: 80.2% >= 70% is S+ on the documented scale
+  // Removed (2026-09-15): minimax/minimax-m3:free (MiniMax M3 (free)) — free variant removed from gateway
+  // Removed (2026-09-15): thinkingmachines/inkling:free (Inkling (free)) — free variant no longer exposed
+  ['liquid/lfm-2.5-2.6b:free', 'LFM2.5-2.6B (free)', 'C', '-', '64k'], // Added (2026-09-15) — verified via live audit
+  ['z-ai/glm-5.2:free', 'GLM 5.2 (free)', 'B+', '-', '32k'], // Added (2026-09-15) — verified via live audit
+  ['inclusionai/ling-3.0-flash-fin:free', 'Ling 3.0 Flash Fin (free)', 'B', '-', '262k'], // Added (2026-09-15) — verified via live audit
+  ['inclusionai/ling-3.0-flash-sante:free', 'Ling 3.0 Flash Sante (free)', 'B', '-', '262k'], // Added (2026-09-15) — verified via live audit
+  ['inclusionai/ling-3.0-flash-vl:free', 'Ling 3.0 Flash VL (free)', 'B+', '-', '262k'], // Added (2026-09-15) — verified via live audit
+  ['nex-agi/nex-n2.5-mini:free', 'Nex AGI Nex-N2.5-Mini (free)', 'B+', '-', '262k'], // Added (2026-09-15) — verified via live audit
+  ['nex-agi/nex-n2.5-pro:free', 'Nex AGI Nex-N2.5-Pro (free)', 'A', '-', '262k'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 LLM7 source - https://api.llm7.io/v1
@@ -436,7 +479,7 @@ export const llm7 = [
   ['minimax-m2.7', 'MiniMax M2.7', 'S+', '78.0%', '180k'],
   // ── A+ tier — SWE-bench Verified 50–60% ──
   // Removed (2026-09-05): gemini-3.1-flash-lite (Gemini 3.1 Flash Lite) — now tier=pro usage_based_only (paid) per live /v1/models
-  ['gpt-oss', 'GPT OSS 20B', 'A+', '50.3%', '131k'], // Fixed (2026-09-05): id 'gpt-oss:20b' → 'gpt-oss', ctx '128k' → '131k' (live 131072)
+  // Removed (2026-09-15): gpt-oss (GPT OSS 20B) — removed from LLM7 API catalog
   ['mistral-Nemo-Instruct-2407', 'Mistral Nemo 12B Instruct', 'A-', '-', '128k'], // Added (2026-08-13)
   // ── A tier — SWE-bench Verified 40–50% ──
   ['codestral-latest', 'Codestral Latest', 'A', '40.0%', '32k'],
@@ -451,7 +494,7 @@ export const routeway = [
   // Removed (2026-09-05): step-3.7-flash:free (Step 3.7 Flash) - free variant discontinued, only paid step-3.7-flash remains ($0.20/$1.15 per M)
   ['minimax-m2.7:free', 'MiniMax M2.7', 'S+', '78.0%', '42k'], // Added (2026-09-02)
   ['muse-glimmer-30b:free', 'Muse Glimmer 30B', 'B+', '-', '131k'], // Added (2026-09-02)
-  ['kimi-k2.6:free', 'Kimi K2.6', 'S+', '-', '42k'],
+  // Removed (2026-09-15): kimi-k2.6:free (Kimi K2.6) — free variant removed, now paid-only; replacement: minimax-m2.7:free
   // ── S tier — SWE-bench Verified 60–70% ──
   // Removed (2026-09-05): laguna-xs.2:free (Poolside Laguna XS.2) - laguna-xs.2 no longer offered in any form, superseded by paid laguna-s-2.1
   // Removed (2026-09-05): gpt-oss-120b:free (GPT OSS 120B) - free variant discontinued, only paid gpt-oss-120b remains ($0.04/$0.30 per M)
@@ -480,6 +523,11 @@ export const novita = [
   // Removed (2026-07-27): qwen/qwen3.5-plus (Qwen3.5 Plus) — no longer in novita catalog
   ['inclusionai/ling-3.0-flash-fin', 'Ling 3.0 Flash Fin', 'B+', '-', '256k'],
   ['inclusionai/ling-3.0-flash-sante', 'Ling 3.0 Flash Sante', 'B+', '-', '256k'],
+  ['bunny', 'Bunny (free tier)', 'C', '-', '256k'], // Added (2026-09-15) — verified via live audit
+  ['qwen/qwen3.6-plus', 'Qwen 3.6 Plus (free tier)', 'A', '-', '1M'], // Added (2026-09-15) — verified via live audit
+  ['qwen/qwen3.5-plus', 'Qwen 3.5 Plus (free tier)', 'A-', '-', '1M'], // Added (2026-09-15) — verified via live audit
+  ['dev/glm46', 'GLM 4.6 (dev, free)', 'A-', '-', '256k'], // Added (2026-09-15) — verified via live audit
+  ['inclusionai/ling-3.0-flash-vl', 'Ling 3.0 Flash VL', 'B+', '-', '256k'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 Pollinations AI source - https://gen.pollinations.ai
@@ -515,7 +563,7 @@ export const pollinations = [
 // 📖 and still reachable with free-tier rate limits (1000 RPM). Keep only the chat text models here.
 export const siliconflow = [
   // ── A tier — SWE-bench Verified 40–50% ──
-  ['THUDM/GLM-Z1-9B-0414', 'GLM-Z1 9B', 'A', '-', '131k'],
+  ['THUDM/GLM-Z1-9B-0414', 'GLM-Z1 9B', 'A', '-', '131k'], // Fixed (2026-09-15): ctx '131k' → '32k'
   ['deepseek-ai/DeepSeek-R1-0528-Qwen3-8B', 'DeepSeek R1 0528 Qwen3 8B', 'A', '-', '131k'],
   // ── B+ tier ──
   ['Qwen/Qwen3-8B', 'Qwen3 8B', 'B+', '30.0%', '131k'],
@@ -523,6 +571,7 @@ export const siliconflow = [
   ['Qwen/Qwen3.5-4B', 'Qwen3.5 4B', 'A-', '-', '262k'],
   ['THUDM/GLM-4-9B-0414', 'GLM-4 9B', 'B+', '-', '32k'],
   ['Qwen/Qwen2.5-7B-Instruct', 'Qwen2.5 7B Instruct', 'B', '-', '32k'],
+  ['Qwen/Qwen2.5-Coder-7B-Instruct', 'Qwen2.5 Coder 7B Instruct', 'B+', '-', '32k'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 Requesty source - https://router.requesty.ai/v1
@@ -563,8 +612,10 @@ export const orcarouter = [
   ['deepseek/deepseek-v4-flash-free', 'DeepSeek V4 Flash (Free)', 'S+', '79.0%', '1M'],
   // ── S tier — SWE-bench Verified 60–70% ──
   ['tencent/hy3-free', 'Tencent Hy3 (Free)', 'S', '-', '256k'],
+  ['orcarouter/free', 'OrcaRouter Free (difficulty-routed)', 'S', '-', '-'], // Added (2026-09-15) — verified via live audit
   // ── A+ tier — SWE-bench Verified 50–60% ──
-  ['qwen/qwen3.8-27b-free', 'Qwen3.8 27B (Free)', 'A+', '-', '64k'],
+  // Removed (2026-09-15): qwen/qwen3.8-27b-free (Qwen3.8 27B (Free)) — no longer in catalog; only paid variant remains; replacement: z-ai/glm-5.3-flash-free
+  ['z-ai/glm-5.3-flash-free', 'GLM-5.3 Flash (Free)', 'A+', '-', '1M'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 Vercel AI Gateway source - https://vercel.com/docs/ai-gateway
@@ -578,11 +629,13 @@ export const orcarouter = [
 // 📖 which is why this provider is quotaCode 'limited'.
 export const vercelGateway = [
   // ── S+ tier — SWE-bench Verified ≥70% ──
-  ['minimax/minimax-m3-free', 'MiniMax M3 (Free)', 'S+', '78.4%', '1M'], // score mirrors minimax-m3 (ollama-cloud, measured 2026-07-27)
-  ['minimax/minimax-m2.7-free', 'MiniMax M2.7 (Free)', 'S+', '78.0%', '196k'], // score mirrors minimax-m2.7 (ollama-cloud, measured 2026-07-27)
+  // Removed (2026-09-15): minimax/minimax-m3-free (MiniMax M3 (Free)) — free variant removed, now paid-only
+  // Removed (2026-09-15): minimax/minimax-m2.7-free (MiniMax M2.7 (Free)) — free variant removed, now paid-only
   ['poolside/laguna-s-2.1-free', 'Laguna S 2.1 (Free)', 'S+', '-', '256k'], // tier follows family precedent: laguna-xs-2.1 ships S+ 70.9% via NVIDIA. Caution: models.dev flags laguna-s-2.1 deprecated (2026-09-05, Zen promo ended) but Vercel still serves the $0 variant live - re-verify at next audit
   // ── B+ tier — vertical-tuned lightweight (coding secondary) ──
   ['inclusionai/ling-3.0-flash-fin-free', 'Ling 3.0 Flash Fin (Free)', 'B+', '-', '256k'], // 124B MoE (5.1B active), finance-tuned, retains coding + math
+  ['inclusionai/ling-3.0-flash-sante-free', 'Ling 3.0 Flash Sante (Free)', 'B+', '-', '256k'], // Added (2026-09-15) — verified via live audit
+  ['inclusionai/ling-3.0-flash-vl-free', 'Ling 3.0 Flash VL (Free)', 'C', '-', '256k'], // Added (2026-09-15) — verified via live audit
 ]
 
 // 📖 Ollama Cloud source - https://ollama.com/pricing and https://ollama.com/search?c=cloud
@@ -600,6 +653,7 @@ export const ollamaCloud = [
   ['deepseek-v4-flash:0731', 'DeepSeek V4 Flash', 'S+', '79.0%', '1M'], // Fixed (2026-08-23): ID 'deepseek-v4-flash' → 'deepseek-v4-flash:0731' (renamed upstream)
   ['deepseek-v4-pro:0813', 'DeepSeek V4 Pro', 'S+', '80.6%', '1M'], // Fixed (2026-08-23): ID 'deepseek-v4-pro' → 'deepseek-v4-pro:0813' (renamed upstream)
   ['glm-5.3', 'GLM 5.3', 'S+', '-', '1M'],
+  ['deepseek-v4.1-flash', 'DeepSeek V4.1 Flash', 'S+', '-', '1M'], // Added (2026-09-15) — verified via live audit
   // ── S tier — SWE-bench Verified 60–70% ──
   ['kimi-k2.7-code', 'Kimi K2.7 Code', 'S', '60.4%', '256k'], // Fixed (2026-07-27): ctx '262k' → '256k'
   ['gpt-oss:120b', 'GPT OSS 120B', 'S', '62.4%', '128k'],