Card text in the human style, two targeted links

README.md

@@ -15,20 +15,17 @@ thumbnail: https://stealthmodels.com/pixel-canary/assets/social-card-v3.png
 
 **Our reasoning study pointed to Qwen two days before Callstack revealed Pixel Canary as Apex, a Qwen-based coding model built for React Native.**
 
-*Independent benchmark card by [Stealth Models](https://stealthmodels.com/?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card): our test results and findings for this model.*
+By [Stealth Models](https://stealthmodels.com/?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card).
 
 <a href="https://stealthmodels.com/benchmarks/models/pixel-canary?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card"><img src="https://huggingface.co/stealth-model/pixel-canary/resolve/main/panels/verdict.webp" alt="StealthMark 1.8.8 results for Pixel Canary (Callstack Apex): benchmark score 49.5 out of 100, General Intelligence 40.4, Visual Intelligence 59.6. Strongest at characters that interact: 76.8 on Raccoon."></a>
 
 <a href="https://stealthmodels.com/benchmarks/models/pixel-canary?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card"><img src="https://huggingface.co/stealth-model/pixel-canary/resolve/main/panels/results.webp" alt="Every StealthMark result of Pixel Canary (Callstack Apex): Raccoon 76.8, Selfie 74.1, Beach 61.4, Mona Lisa 45.9, Ground Truth 44.1, Newton's Mirror 32.7, Latch 100.0 (partial run, 1 of 5 levels). Not tested yet: Foresight."></a>
 
-**Open each result** · Visual Intelligence: [Raccoon&nbsp;76.8](https://stealthmodels.com/benchmarks/models/pixel-canary/raccoon?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card) · [Selfie&nbsp;74.1](https://stealthmodels.com/benchmarks/models/pixel-canary/selfie?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card) · [Beach&nbsp;61.4](https://stealthmodels.com/benchmarks/models/pixel-canary/beach?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card) · [Mona&nbsp;Lisa&nbsp;45.9](https://stealthmodels.com/benchmarks/models/pixel-canary/mona-lisa?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card)<br>
-General Intelligence: [Ground&nbsp;Truth&nbsp;44.1](https://stealthmodels.com/benchmarks/models/pixel-canary/ground-truth?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card) · [Newton's&nbsp;Mirror&nbsp;32.7](https://stealthmodels.com/benchmarks/models/pixel-canary/newtons-mirror?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card) · [Latch&nbsp;100.0&nbsp;(1&nbsp;of&nbsp;5&nbsp;levels)](https://stealthmodels.com/benchmarks/models/pixel-canary?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card#game-latch-five-levels)
-
-**[All results and comparisons on stealthmodels.com ↗](https://stealthmodels.com/benchmarks/models/pixel-canary?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card)**
+[Compare it with the Qwen 3.8 family](https://stealthmodels.com/benchmarks?models=qwen3.8-max-0902,qwen3.8-27b,pixel-canary,qwen3.8-flash,qwen3.6-27b&utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card)
 
 <a href="https://stealthmodels.com/pixel-canary/?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card"><img src="https://huggingface.co/stealth-model/pixel-canary/resolve/main/panels/case.webp" alt="The case file. First seen 25 Sep 2026, revealed 1 Oct 2026, developer Callstack (Poland), context 262,144 tokens, input Text, tool calls unknown, speed in our tests about 255 tok/s. 26 Sep 2026: A tokenizer study by AI時短ラボ links its vocabulary to Qwen. 29 Sep 2026, our finding: Our reasoning study puts Qwen3.8-Flash closest, two days before the reveal. 1 Oct 2026: Callstack reveals Apex: a Qwen foundation with React Native post-training."></a>
 
-Sources: [Callstack's announcement](https://www.callstack.com/blog/apex-agetic-coding-for-mobile-and-web-spcialied-in-react) · [Callstack's model documentation](https://apex.callstack.com/models) · [the AI時短ラボ tokenizer investigation](https://www.ai-jitan-hub.com/news/pixel-canary-stealth-model) · [our reasoning study](https://stealthmodels.com/pixel-canary/?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card)
+[See our reasoning study](https://stealthmodels.com/pixel-canary?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card#reasoning-comparison)
 
 <details><summary>Facts as text</summary>
 
@@ -46,8 +43,8 @@ Sources: [Callstack's announcement](https://www.callstack.com/blog/apex-agetic-c
 </details>
 
 ## Run it
-Callstack serves Apex as a hosted model; no weights are published. If that changes, this card links them.
+Callstack serves Apex as a hosted model, so there are no weights to download.
 
 ---
-[Model guide](https://stealthmodels.com/pixel-canary/?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card) · [StealthMark](https://stealthmodels.com/stealthmark?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card) · [Play the benchmarks](https://stealthmodels.com/play?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card) · [Follow for new reveals](https://huggingface.co/stealth-model)<br>
-This card and its images: CC BY 4.0, credit stealthmodels.com. The model's weights have their own license. Updated 28 October 2026.
+[Model guide](https://stealthmodels.com/pixel-canary/?utm_source=huggingface&utm_medium=referral&utm_campaign=pixel-canary-card) · [StealthMark leaderboard](https://huggingface.co/spaces/stealth-model/stealthmark) · [Follow us](https://huggingface.co/stealth-model)<br>
+Text and images: CC BY 4.0, stealthmodels.com.