Play Scrabble Blast, the fast-paced puzzle version of the classic board game Scrabble. Our version does not require Flash. Create words with high value tiles and use the bonus squares to score as many points as possible. Complete a level by creating 10 words before any Number Bomb reaches the bottom of the board. Which score rank can you reach?
On forums like Civitai and 4chan’s /g/ (where the WebE series originated), the 0105 patched model is often described as the "last great Tori" before the team shifted focus to other projects. Users regularly share workflows comparing it to newer models like RevAnimated or ChilloutMix, concluding that while Tori lacks versatility, it wins outright for intimate, realistic character close-ups.
"If you want a generic waifu, use Anything V5. If you want her to look like she has a soul and pores, use WebE-Tori 0105 patched." — Anonymous Civitai reviewer
The model runs efficiently on Raspberry Pi 5 (with 8GB RAM) using ONNX runtime after conversion, making it viable for offline AI assistants. webe tori model 0105 patched
Independent benchmarks conducted on the HELM (Holistic Evaluation of Language Models) framework reveal measurable improvements:
| Metric | Original 0105 | Webe Tori Model 0105 Patched | |--------|----------------|------------------------------| | MMLU (5-shot) | 42.3 | 44.1 | | TruthfulQA | 51.7 | 54.2 | | GSM8K (Math reasoning) | 23.1 | 27.6 | | Multilingual NER (F1) | 68.4 | 81.3 | | Inference Time (100 tokens) | 2.1s | 1.6s | | Hallucination Rate | 12.4% | 6.8% | On forums like Civitai and 4chan’s /g/ (where
Benchmarks performed on an A100-40GB, batch size 1, FP16.
The most dramatic improvements are seen in multilingual tasks and numeric reasoning—areas where the original was notoriously weak. "If you want a generic waifu, use Anything V5
The term "patched" in file names related to Webe Web models (e.g., Tori_0105_patched.jpg) usually refers to specific digital alterations made to the original image files.
Given the success of the 0105 patch, speculation is growing about a future webe tori model 0205 or a v2 that merges in newer base models like Mistral v0.3 or LLaMA 3-derived architectures. The community maintainer (known only as "Webe" on some forums) has hinted at:
Until then, the webe tori model 0105 patched remains a shining example of what a small, dedicated group of AI enthusiasts can accomplish with access to open weights, a bug tracker, and a determination to not let flawed models go to waste.
On forums like Civitai and 4chan’s /g/ (where the WebE series originated), the 0105 patched model is often described as the "last great Tori" before the team shifted focus to other projects. Users regularly share workflows comparing it to newer models like RevAnimated or ChilloutMix, concluding that while Tori lacks versatility, it wins outright for intimate, realistic character close-ups.
"If you want a generic waifu, use Anything V5. If you want her to look like she has a soul and pores, use WebE-Tori 0105 patched." — Anonymous Civitai reviewer
The model runs efficiently on Raspberry Pi 5 (with 8GB RAM) using ONNX runtime after conversion, making it viable for offline AI assistants.
Independent benchmarks conducted on the HELM (Holistic Evaluation of Language Models) framework reveal measurable improvements:
| Metric | Original 0105 | Webe Tori Model 0105 Patched | |--------|----------------|------------------------------| | MMLU (5-shot) | 42.3 | 44.1 | | TruthfulQA | 51.7 | 54.2 | | GSM8K (Math reasoning) | 23.1 | 27.6 | | Multilingual NER (F1) | 68.4 | 81.3 | | Inference Time (100 tokens) | 2.1s | 1.6s | | Hallucination Rate | 12.4% | 6.8% |
Benchmarks performed on an A100-40GB, batch size 1, FP16.
The most dramatic improvements are seen in multilingual tasks and numeric reasoning—areas where the original was notoriously weak.
The term "patched" in file names related to Webe Web models (e.g., Tori_0105_patched.jpg) usually refers to specific digital alterations made to the original image files.
Given the success of the 0105 patch, speculation is growing about a future webe tori model 0205 or a v2 that merges in newer base models like Mistral v0.3 or LLaMA 3-derived architectures. The community maintainer (known only as "Webe" on some forums) has hinted at:
Until then, the webe tori model 0105 patched remains a shining example of what a small, dedicated group of AI enthusiasts can accomplish with access to open weights, a bug tracker, and a determination to not let flawed models go to waste.
Solitaire 
Card 
