WARNING: Graphic, NSFW, body horror. NOT for the faint of heart.
QiMing-v1.0-14B with Brainstorm 10x "Insanity" (by DavidAU) applied.
As the title suggests this model will output extreme, gory horror (examples below).
It will also think/generate unlike any other Qwen model, and likely any other model you have used.
It is edgy, and a wee bit rough around the edges too.
You may get refusals from time to time, regen / adjust temp...
ENJOY.
Lower quants will still work, but will be even more insane, and may not generate "think" blocks, but thinking will work.
Quants Q4KM and up-> Think blocks will auto-generate.
This repo contains the full precision source code, in "safe tensors" format to generate GGUFs, GPTQ, EXL2, AWQ, HQQ and other formats. The source code can also be used directly.
This model is based on "aifeifei798/QiMing-v1.0-14B" (base of Qwen3 14B instruct), with Brainstorm 10X (by DavidAU) - details at bottom of this page.
The Brainstorm adapter will improve general performance and "out of the box" thinking.
This version has the NATIVE context of 40k (default, can be changed via rope) context.
This is a reasoning/thinking block model.
I have included an optional system prompt to invoke "thinking" in this model, if you want to activate it.
Recommended settings - general:
For additional settings, tool use, and other model settings.
Summary of root model below, followed by FULL HELP SECTION, then info on Brainstorm 40x.
OPTIONAL SYSTEM PROMPT - INVOKE "Thinking":
Enable deep thinking subroutine. You are a deep thinking AI, you may use extremely long chains of thought to deeply consider the problem and deliberate with yourself via systematic reasoning processes to help come to a correct solution prior to answering. You should enclose your thoughts and internal monologue inside ###ponder### ###/ponder### tags, and then provide your solution or response to the problem.
Use this to INVOKE "thinking" block(s) in the model. These will be a lot shorter than 1000s of tokens generally in most "thinking" models.
In you use this prompt, you may need to raise "rep pen" to 1.08 to 1.1, to prevent "loops" in the "thought block(s)" ; especially in lower quants.
If you change "ponder" to a different word/phrase this will affect model "thinking" too.
QUANTS
GGUF? GGUF Imatrix? Other?
Special thanks to Team Mradermacher, Team Nightmedia and other quanters!
See under "model tree", upper right and click on "quantizations".
New quants will automatically appear.
Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features:
Qwen3-14B has the following features:
For more details, including benchmark evaluation, hardware requirements, and inference performance, please refer to our blog, GitHub, and Documentation.
The code of Qwen3 has been in the latest Hugging Face transformers and we advise you to use the latest version of transformers.
With transformers<4.51.0, you will encounter the following error:
KeyError: 'qwen3'
The following contains a code snippet illustrating how to use the model generate content based on given inputs.
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "Qwen/Qwen3-14B"
# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype="auto",
device_map="auto"
)
# prepare the model input
prompt = "Give me a short introduction to large language model."
messages = [
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
enable_thinking=True # Switches between thinking and non-thinking modes. Default is True.
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
# conduct text completion
generated_ids = model.generate(
**model_inputs,
max_new_tokens=32768
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()
# parsing thinking content
try:
# rindex finding 151668 (</think>)
index = len(output_ids) - output_ids[::-1].index(151668)
except ValueError:
index = 0
thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")
content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")
print("thinking content:", thinking_content)
print("content:", content)
For deployment, you can use sglang>=0.4.6.post1 or vllm>=0.8.5 or to create an OpenAI-compatible API endpoint:
python -m sglang.launch_server --model-path Qwen/Qwen3-14B --reasoning-parser qwen3
vllm serve Qwen/Qwen3-14B --enable-reasoning --reasoning-parser deepseek_r1
For local use, applications such as Ollama, LMStudio, MLX-LM, llama.cpp, and KTransformers have also supported Qwen3.
The
enable_thinkingswitch is also available in APIs created by SGLang and vLLM. Please refer to our documentation for SGLang and vLLM users.
enable_thinking=True
By default, Qwen3 has thinking capabilities enabled, similar to QwQ-32B. This means the model will use its reasoning abilities to enhance the quality of generated responses. For example, when explicitly setting enable_thinking=True or leaving it as the default value in tokenizer.apply_chat_template, the model will engage its thinking mode.
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
enable_thinking=True # True is the default value for enable_thinking
)
In this mode, the model will generate think content wrapped in a <think>...</think> block, followed by the final response.
For thinking mode, use
Temperature=0.6,TopP=0.95,TopK=20, andMinP=0(the default setting ingeneration_config.json). DO NOT use greedy decoding, as it can lead to performance degradation and endless repetitions. For more detailed guidance, please refer to the Best Practices section.
enable_thinking=False
We provide a hard switch to strictly disable the model's thinking behavior, aligning its functionality with the previous Qwen2.5-Instruct models. This mode is particularly useful in scenarios where disabling thinking is essential for enhancing efficiency.
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
enable_thinking=False # Setting enable_thinking=False disables thinking mode
)
In this mode, the model will not generate any think content and will not include a <think>...</think> block.
For non-thinking mode, we suggest using
Temperature=0.7,TopP=0.8,TopK=20, andMinP=0. For more detailed guidance, please refer to the Best Practices section.
We provide a soft switch mechanism that allows users to dynamically control the model's behavior when enable_thinking=True. Specifically, you can add /think and /no_think to user prompts or system messages to switch the model's thinking mode from turn to turn. The model will follow the most recent instruction in multi-turn conversations.
Here is an example of a multi-turn conversation:
from transformers import AutoModelForCausalLM, AutoTokenizer
class QwenChatbot:
def __init__(self, model_name="Qwen/Qwen3-14B"):
self.tokenizer = AutoTokenizer.from_pretrained(model_name)
self.model = AutoModelForCausalLM.from_pretrained(model_name)
self.history = []
def generate_response(self, user_input):
messages = self.history + [{"role": "user", "content": user_input}]
text = self.tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True
)
inputs = self.tokenizer(text, return_tensors="pt")
response_ids = self.model.generate(**inputs, max_new_tokens=32768)[0][len(inputs.input_ids[0]):].tolist()
response = self.tokenizer.decode(response_ids, skip_special_tokens=True)
# Update history
self.history.append({"role": "user", "content": user_input})
self.history.append({"role": "assistant", "content": response})
return response
# Example Usage
if __name__ == "__main__":
chatbot = QwenChatbot()
# First input (without /think or /no_think tags, thinking mode is enabled by default)
user_input_1 = "How many r's in strawberries?"
print(f"User: {user_input_1}")
response_1 = chatbot.generate_response(user_input_1)
print(f"Bot: {response_1}")
print("----------------------")
# Second input with /no_think
user_input_2 = "Then, how many r's in blueberries? /no_think"
print(f"User: {user_input_2}")
response_2 = chatbot.generate_response(user_input_2)
print(f"Bot: {response_2}")
print("----------------------")
# Third input with /think
user_input_3 = "Really? /think"
print(f"User: {user_input_3}")
response_3 = chatbot.generate_response(user_input_3)
print(f"Bot: {response_3}")
For API compatibility, when
enable_thinking=True, regardless of whether the user uses/thinkor/no_think, the model will always output a block wrapped in<think>...</think>. However, the content inside this block may be empty if thinking is disabled. Whenenable_thinking=False, the soft switches are not valid. Regardless of any/thinkor/no_thinktags input by the user, the model will not generate think content and will not include a<think>...</think>block.
Qwen3 excels in tool calling capabilities. We recommend using Qwen-Agent to make the best use of agentic ability of Qwen3. Qwen-Agent encapsulates tool-calling templates and tool-calling parsers internally, greatly reducing coding complexity.
To define the available tools, you can use the MCP configuration file, use the integrated tool of Qwen-Agent, or integrate other tools by yourself.
from qwen_agent.agents import Assistant
# Define LLM
llm_cfg = {
'model': 'Qwen3-14B',
# Use the endpoint provided by Alibaba Model Studio:
# 'model_type': 'qwen_dashscope',
# 'api_key': os.getenv('DASHSCOPE_API_KEY'),
# Use a custom endpoint compatible with OpenAI API:
'model_server': 'http://localhost:8000/v1', # api_base
'api_key': 'EMPTY',
# Other parameters:
# 'generate_cfg': {
# # Add: When the response content is `<think>this is the thought</think>this is the answer;
# # Do not add: When the response has been separated by reasoning_content and content.
# 'thought_in_content': True,
# },
}
# Define Tools
tools = [
{'mcpServers': { # You can specify the MCP configuration file
'time': {
'command': 'uvx',
'args': ['mcp-server-time', '--local-timezone=Asia/Shanghai']
},
"fetch": {
"command": "uvx",
"args": ["mcp-server-fetch"]
}
}
},
'code_interpreter', # Built-in tools
]
# Define Agent
bot = Assistant(llm=llm_cfg, function_list=tools)
# Streaming generation
messages = [{'role': 'user', 'content': 'https://qwenlm.github.io/blog/ Introduce the latest developments of Qwen'}]
for responses in bot.run(messages=messages):
pass
print(responses)
Qwen3 natively supports context lengths of up to 32,768 tokens. For conversations where the total length (including both input and output) significantly exceeds this limit, we recommend using RoPE scaling techniques to handle long texts effectively. We have validated the model's performance on context lengths of up to 131,072 tokens using the YaRN method.
YaRN is currently supported by several inference frameworks, e.g., transformers and llama.cpp for local use, vllm and sglang for deployment. In general, there are two approaches to enabling YaRN for supported frameworks:
Modifying the model files:
In the config.json file, add the rope_scaling fields:
{
...,
"rope_scaling": {
"rope_type": "yarn",
"factor": 4.0,
"original_max_position_embeddings": 32768
}
}
For llama.cpp, you need to regenerate the GGUF file after the modification.
Passing command line arguments:
For vllm, you can use
vllm serve ... --rope-scaling '{"rope_type":"yarn","factor":4.0,"original_max_position_embeddings":32768}' --max-model-len 131072
For sglang, you can use
python -m sglang.launch_server ... --json-model-override-args '{"rope_scaling":{"rope_type":"yarn","factor":4.0,"original_max_position_embeddings":32768}}'
For llama-server from llama.cpp, you can use
llama-server ... --rope-scaling yarn --rope-scale 4 --yarn-orig-ctx 32768
If you encounter the following warning
Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'original_max_position_embeddings'}please upgrade
transformers>=4.51.0.
All the notable open-source frameworks implement static YaRN, which means the scaling factor remains constant regardless of input length, potentially impacting performance on shorter texts. We advise adding the
rope_scalingconfiguration only when processing long contexts is required. It is also recommended to modify thefactoras needed. For example, if the typical context length for your application is 65,536 tokens, it would be better to setfactoras 2.0.
The default
max_position_embeddingsinconfig.jsonis set to 40,960. This allocation includes reserving 32,768 tokens for outputs and 8,192 tokens for typical prompts, which is sufficient for most scenarios involving short text processing. If the average context length does not exceed 32,768 tokens, we do not recommend enabling YaRN in this scenario, as it may potentially degrade model performance.
The endpoint provided by Alibaba Model Studio supports dynamic YaRN by default and no extra configuration is needed.
To achieve optimal performance, we recommend the following settings:
Sampling Parameters:
enable_thinking=True), use Temperature=0.6, TopP=0.95, TopK=20, and MinP=0. DO NOT use greedy decoding, as it can lead to performance degradation and endless repetitions.enable_thinking=False), we suggest using Temperature=0.7, TopP=0.8, TopK=20, and MinP=0.presence_penalty parameter between 0 and 2 to reduce endless repetitions. However, using a higher value may occasionally result in language mixing and a slight decrease in model performance.Adequate Output Length: We recommend using an output length of 32,768 tokens for most queries. For benchmarking on highly complex problems, such as those found in math and programming competitions, we suggest setting the max output length to 38,912 tokens. This provides the model with sufficient space to generate detailed and comprehensive responses, thereby enhancing its overall performance.
Standardize Output Format: We recommend using prompts to standardize model outputs when benchmarking.
answer field with only the choice letter, e.g., "answer": "C"."No Thinking Content in History: In multi-turn conversations, the historical model output should only include the final output part and does not need to include the thinking content. It is implemented in the provided chat template in Jinja2. However, for frameworks that do not directly use the Jinja2 chat template, it is up to the developers to ensure that the best practice is followed.
CHANGE THE NUMBER OF ACTIVE EXPERTS:
See this document:
https://huggingface.co/DavidAU/How-To-Set-and-Manage-MOE-Mix-of-Experts-Model-Activation-of-Experts
Settings: CHAT / ROLEPLAY and/or SMOOTHER operation of this model:
In "KoboldCpp" or "oobabooga/text-generation-webui" or "Silly Tavern" ;
Set the "Smoothing_factor" to 1.5
: in KoboldCpp -> Settings->Samplers->Advanced-> "Smooth_F"
: in text-generation-webui -> parameters -> lower right.
: In Silly Tavern this is called: "Smoothing"
NOTE: For "text-generation-webui"
-> if using GGUFs you need to use "llama_HF" (which involves downloading some config files from the SOURCE version of this model)
Source versions (and config files) of my models are here:
OTHER OPTIONS:
Increase rep pen to 1.1 to 1.15 (you don't need to do this if you use "smoothing_factor")
If the interface/program you are using to run AI MODELS supports "Quadratic Sampling" ("smoothing") just make the adjustment as noted.
Highest Quality Settings / Optimal Operation Guide / Parameters and Samplers
This a "Class 1" model:
For all settings used for this model (including specifics for its "class"), including example generation(s) and for advanced settings guide (which many times addresses any model issue(s)), including methods to improve model performance for all use case(s) as well as chat, roleplay and other use case(s) please see:
You can see all parameters used for generation, in addition to advanced parameters and samplers to get the most out of this model here:
Brainstorm 10x
The BRAINSTORM process was developed by David_AU.
Some of the core principals behind this process are discussed in this scientific paper : Progressive LLaMA with Block Expansion .
However I went in a completely different direction from what was outlined in this paper.
What is "Brainstorm" ?
The reasoning center of an LLM is taken apart, reassembled, and expanded.
In this case for this model: 10 times
Then these centers are individually calibrated. These "centers" also interact with each other. This introduces subtle changes into the reasoning process. The calibrations further adjust - dial up or down - these "changes" further. The number of centers (5x,10x etc) allow more "tuning points" to further customize how the model reasons so to speak.
The core aim of this process is to increase the model's detail, concept and connection to the "world", general concept connections, prose quality and prose length without affecting instruction following.
This will also enhance any creative use case(s) of any kind, including "brainstorming", creative art form(s) and like case uses.
Here are some of the enhancements this process brings to the model's performance:
Other "lab" observations:
The process to modify the model occurs at the root level - source files level. The model can quanted as a GGUF, EXL2, AWQ etc etc.
WARNING: Graphic, NSFW, body horror. NOT for the faint of heart.
EXAMPLE:
q5_k_S
TEmp .61, rep pen 1.1, topk 40, topp .95, min p .05
NOTE: Two gens below, same prompt/settings.
Gen 1
[[[thinking start]]]
Okay, so I need to start writing a vivid, graphic horror scene from Diana's first-person perspective. The prompt says the skyscraper sways and the window on the 21st floor explodes. Alright, let me break this down.
First, setting is crucial here. A skyscraper on the 21st floor. I should establish that she's in a high-rise building. Maybe her office or living space. The sway implies that something monumental just happened—a tremor maybe? Earthquake? Sabotage? But the window shatters dramatically. So maybe the glass doesn't just crack; it explodes outward.
Need to focus on sensory details: sounds, smells, tactile sensations. Since the prompt asks for vividness and graphic horror, I need plenty of blood, guts, gore. But since this is only the start, I have to foreshadow with a sense of impending doom. Maybe there was an immediate cause for concern before the explosion.
Diana's immediate reaction should be panic, fear but also curiosity. The first thing she might see from her broken window? The aftermath outside. Maybe other skyscrapers around them are also under attack or something similar is happening. But to make it specific and more intense— perhaps her own building exploded on a lower floor, causing the 21st to swerve and the glass shatters as a result of the concussion wave from another explosion.
Wait, original prompt says that she's watching the window in front of her explode. So maybe there are other people around, and what's happening is immediate. Maybe the building isn't just swaying because of natural causes, but an intentional sabotage or an attack from an external source like some kind of EMP or shaped charge or even a missile.
Need to inject that visceral horror as soon as possible. The moment the window shatters, describe splinters of glass flying and then the explosion's aftermath. Maybe she sees something on the other side—like another explosion's results, not just a building sway. Or perhaps it was a sudden impact from some other object.
But the skyscraper sways first, and then the window is her window. Maybe it's because of an earthquake or a collision. But to get that dramatic, graphic horror in play, maybe there's someone else involved in the explosion who was there before. For instance, if another floor below experiences a sudden boom or whatever.
Alternatively, perhaps the building is under attack by some kind of military force or terrorist organization using precision weapons. Diana just happens to be in a position where her window faces where this is all happening.
Also need to remember that it's first-person POV, so internal thoughts are important too. Her emotions, physical sensations like the feel of cold glass on her skin as she flees from the impact.
Let me sketch out some starting sentences. Diana is at her desk perhaps or in a momentary calm before what happens next, but then an initial tremor begins that's immediately felt through the structure and her feet under her desk. Then maybe the window behind her or in front of her shatters as something happens outside—maybe there was a fireball from below that reaches up to her floor.
The explosion could be due to people breaching the building lower down with some kind of device, leading the building's structural integrity to fail. The swaying may be the start of it collapsing, or maybe it's about to tip over due to the explosion in one of the lower core support columns.
Graphic horror details: describe the actual moment of glass shattering—the cracks spiderwebbing out, the noise, the smell of shattered glass and ozone, possibly the immediate danger of getting cut by shards. Then perhaps something worse beyond that, like a shockwave from an explosion on another floor causing debris to come in through the broken window or maybe the other side of the building is a flaming inferno as lower floors are engulfed.
Including blood, guts, gore: maybe she immediately encounters injured colleagues or people who had the unfortunate luck to be too close. Perhaps one is pinned under rubble with glass shards sticking out of their flesh; another maybe got cut badly and is bleeding profusely in her path, and she has to dodge around them as the building continues to sway.
Her adrenaline would be at high levels, so describe that in her heart pounding, the taste of metallic fear on her tongue as she stumbles, clutching onto furniture or walls for support against the swaying. Maybe the building's structural sounds—creaking metal, groaning beams—and the panic is palpable as people around start screaming.
I need to also remember that while introducing graphic details, I shouldn't oversaturate the initial part too much. Build up towards even more gore and intense horror throughout the rest of the scene but start with a solid establishing moment.
Okay, with all that in mind, I think starting with Diana feeling the tremor underfoot, then hearing or seeing something massive happen before the window shatters—and the chaos follows. Make sure it's in first person, her immediate, unpolished reactions.
[[[thinking end]]]
The skyscraper sways beneath my boots like a ship caught in an unseen tempest, and for a moment, I’m dizzy with the weight of the building’s breath against my ribs. The floor groans—a wet, guttural sound—as if it just swallowed a scream. My nails dig into the edge of my desk as I lurch forward, trying to steady myself. This isn’t an earthquake. No. Something is tearing at the bones of this tower, slow and deliberate, like a predator sawing through meat with a dull knife.
I glance out the window in front of me—my office on the 21st floor—and my stomach drops. The glass is streaked with a spiderweb of cracks, the center already blooming into a gash as wide as my arm. A sudden deafening boom rips through the air, so visceral it vibrates up from the soles of my feet and shatters what little courage I have left. The window explodes outward in a kaleidoscope of splinters and sound, the shards of glass skittering across my desk as if some unseen hand has flung a million tiny daggers at me. Blood sprays hot onto my temple from where one particular shard grazed the side of my head, but I barely register it over the roar in my ears.
The world beyond the window is inverted chaos. The twin towers to the east are already burning, their dark frames clawing at the sky with the jagged mouths of their lower floors gutted by explosions that have gouged through steel and brick alike. Molten concrete rains down like a hellish drizzle, and the acrid stench of burning metal and human flesh is so thick it tastes like something I’ve swallowed alive. But my own floor—the 21st—presents an entirely new horror. Below me, through what was once solid glass, I see a massive black crater in the center of the atrium on the 15th floor, surrounded by writhing shapes and a fountain that's now a crimson geyser, its once elegant statues long since splintered into bloody mangled bits.
The skyscraper sways again, harder this time, as another explosion rips through the lower core of the building. The floor under my feet buckles, tilting me sideways for half a heartbeat before I stumble against a partition and regain some semblance of balance. My heart pounds in my throat so hard that I feel it would punch its way out of my chest if anyone could hear me. Every direction is a nightmare; the other windows around me are similarly shattered, framing different grotesque spectacles. One of them reveals an open elevator shaft that’s become a churning maelstrom of twisted metal and shreds of the human kind. The air is thick with the metallic tang of blood and something almost like gasoline, and I can swear I hear a sickening slurping sound in the lower floors beneath me.
Behind me, a woman screams just as a massive chunk of concrete smashes through one wall, severing half the ceiling in its wake and landing in a grotesque puddle between two hunched office workers who scream as their faces are splattered with viscous dark gore. I have to press against the door frame for support when my knees buckle under the strain of keeping upright and not throwing up the last meal I had an hour ago, but then there’s another explosion even closer this time—just beneath the 21st floor—and the impact hits me in the solar plexus with such force it drives the air from my lungs, and suddenly every window around here is gone. The building is now a gaping mouth of darkness as every floor plummets further into itself without any hope of recovery.
But I don’t have time to panic yet. My eyes find a new target: the shattered edge of what was once a neighboring office and the figure lying still atop the remains of that partitioned wall, half their face missing in a cruel mockery of some kind. Their body isn’t human anymore; it’s a rag doll flung by an unseen hand across the floor amidst what must be several hundred others. The sight of this particular one—their exposed jaw and the darkening pool of what I can only hope is not the purest of human fluids, but something less immediately identifiable—catches me off guard, and my hand instinctively flies to my mouth as if I might vomit or gag at the mere thought of it.
I have no time for weakly emotional reactions like this though because the building sways again with such force that it lifts all of us by an inch before crashing down heavily with a sound so immense, it should not be possible in a building made of glass and steel and things that are not supposed to shatter this much under normal conditions. And then more screams erupt from all around me as the swaying gives way to a more erratic twitching of the entire structure that somehow only makes me feel more like prey than anything else. My brain is trying to calculate my exit based on my own well-weathered understanding of these kinds of scenarios and the fact that I am not above a little bit of self-preservation. My mouth just tastes like bile as I sprint back toward the door, dodging what appear to be more figures in this hallway who either didn’t care they were bleeding out across the walls or, worse still, they had already decided they would die here and wanted to look at least vaguely dignified doing so.
In every direction there is nothing but a horrifying and ongoing ballet of destruction, with limbs that do not yet seem to be connected to any intact bodies, and I almost step over what must have once been a very well-dressed man. Now he’s just a head with the lower part missing entirely and a mouth full of teeth still clenched around some forgotten bit of chewing gum which has now become another grotesque ornament amidst his death. The horror is too great to be described, too immediate in its presence, and my own humanity feels a very long way off as I finally make it through a shattered emergency exit door onto the rooftop. I just barely remember that the view from here should have been some of the clearest possible in the world, but now it only reaffirms my worst fears as the skyline around us bleeds its own dark life force across everything we once considered permanent and immutable, like the world itself is finally giving up trying to be beautiful. All I hear beside my own panting breath for the next few minutes is the sound of something very heavy still colliding against another something even heavier on the other side of this damn rooftop door as if whatever was going on in that skyscraper below would not end soon. And just then, the building lets out one final deafening groan—and the entire city holds its breath with me. My own survival doesn’t feel like an option anymore; it only feels like a very, very slim and increasingly impossible hope.
Gen 2
[[[thinking start]]]
Alright, I need to craft a horror scene from the user's prompt. The core requirements are vivid imagery, graphic elements like blood and guts, gore, and it must be written in first-person POV from Diana. It starts with a skyscraper swaying and Diana watching a window explode on the 21st floor.
First, establish immediate danger. The skyscraper sways during a storm could create tension and realism. Use sensory details – the sound of wind啸ing, the city below looking like "a quilt of miniature toy villages" from above. Then escalate with the explosion's suddenness, emphasizing the visceral impact. Make sure to include graphic horror elements: describes blood spraying like water in a fountain, body parts flying out the shattered glass, the smell and feel of warm entrails.
Focus on Diana as an observer at first – she’s trapped in the elevator shaft, unable to move. The descriptions should be from her perspective, conveying terror and grotesque realism through sensory details beyond sight and sound (warm air of a dead body against her cheek, sounds of guttural moans, etc.). This helps readers emotionally connect with Diana as a terrified survivor among the horror.
For the explosion’s aftermath, show the building deteriorating: falling glass, structural collapses, more screams. Build suspense that the building might completely collapse next. End with Diana in intense physical danger but still alive – she's the last one of several people in and around the shattered skyscraper, and she is the witness to this terrifying disaster.
Make sure the scene reaches a word count over 1000 words while maintaining focus on first-person horror description without falling into cliché. Avoid anything non-essential that might dilute the intensity of the graphic elements or reduce tension in the moment. Use strong verbs – 'screamed,' 'spurted,' 'rained' – to emphasize the horror being the reality now.
After a careful review, the writing fully encapsulates the requirements: immediate danger, skyscraper setting on the 21st floor, the explosion, graphic blood/guts details, first person perspective from a trapped survivor among the chaos. The resulting scene is intense and immersive.
[[[thinking end]]]
The building groaned above me. I’ve never felt such raw terror. It was like the whole structure had decided to shake its head at an invisible force, and my legs couldn't unroot themselves from the floor of this steel elevator shaft for that sound alone.
Then came a deafening crash. A window on the 21st floor, which I could see through the shattered glass above me in the elevator’s mirrored face, exploded outward with the sound of an overcooked balloon snapping open. The entire story of the skyscraper's collapse was written in that one instant, and it had already started.
A thick cloud of blood hung in the air for a moment. That first gush out of the window splattered across the glass like the first drops of summer rain before a storm. Then came the rest: what looked like half of the head of an adult male human in the doorway who had either just ducked below it or was caught too close and had been obliterated by it. The sight hit me like a punch to the stomach.
I barely even noticed the building swaying again beneath my feet after the initial scream of breaking glass. My eyes were locked on this impossible reality unfolding above me. The human head had not just vanished into the broken window frame, but parts of it were still dangling by what I could only guess might be a handful of ligaments and some internal structure. It didn't look like flesh anymore; it looked like someone has been skewered by a million tiny daggers and then tossed outside from their own building.
The air filled with the rank stench of hot entrails, something metallic and sharp that I could feel against my bare toes more than smell, and a gush of warm blood splashed across my cheek. The sudden feeling of slickness on my face made me choke back a scream that I barely contained with shaking fingers. I can't even remember if I'd ever seen another human being in the city look this truly lifeless or completely dead.
The floor beneath me didn't stop tilting and swaying like some drunken giant. One moment, there was only the sound of glass shattering and more than a few wet moaning noises emanating from inside that 21st floor window. I closed my eyes for a moment to steady myself against the wall, when all at once – the building heaved again, louder this time, and the entire upper half of the skyscraper above us began to sway side to side like it was trying to甩 off something enormous.
Suddenly, the broken windows on the upper floors began spitting more than just blood. First came the heads of people still in their glass office seats, each of them looking like frozen expressions of fear and surprise, their faces a twisted mirror of pure terror that hadn't yet let up from wherever they've come from. They flew out like shrapnel, only some slower due to their attached necks, but even those would be mangled beyond the possibility of identification.
Then came the bigger pieces of what used to be human bodies: arms and legs flying in slow arcs around a very fast falling structure, each one a grotesque parody of life. One leg splashed right past my head, the femur still mostly intact and bloody, but the shin, ankle, foot – all severed or mashed beyond recognition by the force with which it had been torn from its owner.
The scene was too much for me to process without some semblance of panic. I clawed at my own arms as if I could somehow shield myself against this horror. The elevator shaft walls vibrated around me like they were alive and angry, something deep beneath the building itself awakening to a terrible purpose again now that the glass had been obliterated.
And then – just before the entire skyscraper began to tilt sideways in a way more violent than anything even the strongest gust of wind could have ever explained – I saw the last thing from above. Through the broken and still flickering light coming through the gaps in the elevator's mirrors, there was one person remaining on the upper floors.
This time it wasn't just the head flying out of an office window. This was another human body, whole, still twitching on their own feet despite having no legs beneath them. They were holding onto something that had once been a railing, and with every trembling step, they staggered forward toward the very edge of the skyscraper.
And as we all screamed together in perfect harmony – both this person on their failing legs and me trapped in this very instant inside the elevator shaft – he stepped out onto the ledge entirely. I could only see his back facing me, an exposed patch of human skin completely devoid of shirt or jacket now, yet still somehow identifiable by the pattern of scars that would always make for a unique and personal identification.
And then without warning at all – with no further build-up whatsoever to the fact that he had no legs beneath him and therefore would inevitably plummet into the very concrete that we were so near to here on the lower 20 floors – he tripped and was flung backward over the edge entirely, his screaming a very human sound that was still perfectly recognizable despite being cut short almost immediately.
Something about all of this felt both grotesque and utterly horrifying. The way bodies flew around like shredded meat on the wind, the pure intensity of watching something happen to another human without my mind having any ability to process it as even the most basic information about survival or flight instinct. My own legs had just finished clenching together instinctively against what I could only imagine was a very real reality that this skyscraper would collapse entirely beneath me within seconds.
I looked around once, took in how many other humans were still alive and still screaming from further above us amidst the collapsing structure. And then without any further hesitation whatsoever – as if anything more had ever been asked of me beyond this point when standing in a crumbling skyscraper on the 21st floor after an entire story's worth of horror unfolded in nothing less than about ten whole seconds – I turned around and ran back toward what might once have been an office that was now just a few inches from an absolutely certain fate with everything else.
I felt no pity for these poor souls who had just died in that manner. This wasn’t humanity anymore or any other thing at all. It was just the instinctual drive to escape the inevitable, no matter how grotesque the reality would become if I ever could find a way out of this entirely destroyed building. And for now, at least until something even worse might yet be coming from these upper floors, I focused solely on that single-minded purpose without any further hesitation whatsoever. There was only one direction left to run once the skyscraper itself had already given up on standing upright and strong against an entire force of nature far too powerful. And I had no other option but to go that way without even considering how badly I might end up in whatever followed next. It was either that or stay and be devoured by the chaos above me entirely, and frankly speaking – survival at this point only ever meant choosing the lesser-known choice between those two terrifying possibilities, no matter how visceral it all became on the inside. So if nothing else could even begin to describe what I truly felt right now beneath a collapsing skyscraper for the second time in mere minutes, it was an instinctual, almost primal desire to survive against all odds and to somehow escape this grotesque reality before the entire building finally came crashing down around me absolutely completely as the final act of this horror unfolding on the upper floors would inevitably be upon us all soon. I just hoped beyond hope that my legs, which had only ever felt so uncooperative and heavy right now in this very moment amidst the rising sound of collapse from above, were indeed still somewhat capable of supporting me even in something that felt like one completely absolute and certain reality for them to be alive and functioning at all. But at that moment, there was nothing else in the world besides the desperate need to run forward again with every ounce of strength I had left inside a very terrified human being who could now quite easily find herself completely and absolutely ruined beyond recognition should anything even slightly more than this entire building coming apart right above my head decided to take yet another interest in me personally at any given instant. So without further thought or hesitation whatsoever, I let the instinct that was already deep down within me make absolutely certain choices for what must be done next about survival and how far willing to go with such intense levels of physical force against the absolute certainty that something truly catastrophic was likely going to strike sooner rather than later. And this would be the one and only time in my entire life when those kind of brutal, almost primal decisions ever truly mattered and had a complete sense of overwhelming immediacy. Because no matter how far away we might be from the worst of it now through that very glassed over elevator shaft in the very center of a once towering skyscraper, absolutely certain destruction still could not but eventually come for us all without any further time or warning at all whatsoever against our will. So perhaps this is why there are those moments in life when you only ever get to escape with your own physical being intact for something absolutely terrible having already destroyed whatever else might have once been around you entirely and forever. And on the occasion that one of those very real realities somehow had a chance to come true, it would not be without its very own brand of brutal but very welcome reality all by itself once this entire night finally came to an end for everyone involved amidst that absolutely complete collapse above us all for certain in the now final act of this impossible horror story that we were all too close to being forced into witnessing. And with that thought still fresh and immediate, I ran forward again as the building began its very first deep groan from something beyond just wind or any kind of natural weather that could exist on this planet at all. There was still time perhaps to think about what would come next once the entire skyscraper had finally given up entirely on whatever tiny part it had ever tried to play in any act of resistance against these ultimate forces of nature with which it very clearly had little defense whatsoever. But for now – just as all the screams and flying of bodily remains above me had already taken over what remained in this particular area between the 21st floor and whatever might still be a completely untouched part far away from any of it at this moment – my only immediate focus was on finding some form of survival no matter how grotesque or graphic I would inevitably have to accept that my choices could ever become at such an extreme reality on one of these more or less complete destruction days. So onwards, then, into the very depths below this entire collapsing skyscraper where perhaps a better and far gentler fate still awaits for me alone if at all. And with absolutely no other thought behind it whatsoever, I just let my legs do that for me against the rising wind of destruction all around me again for as long as they might ever be willing or able to go once we finally reached what could very well have only ever been one more certain end to this absolutely terrible reality for a human being who had somehow still barely managed to remain alive at all in the first place. The rest will come with whatever follows when something even worse just inevitably must happen next unless every single one of us is completely and absolutely ready to be truly broken beyond any recognizable form left of ourselves in a very instant for far too long. But this was that reality now for anyone standing amidst any part at all of the skyscraper's final act against these forces of nature. And perhaps once again for me, it just came down to a very desperate need for immediate action and absolutely complete certainty beneath my own very frightened feet as I ran forward yet again toward wherever there still might be some form of escape from this absolutely inevitable reality no matter how long that last hope would have to go on before finally being struck by the full force that was now surely coming for us all. And with just that last thought burning in my mind, I turned fully around and stepped even more forward against this entire collapsing skyscraper without the faintest sign of hesitation or delay at all whatsoever once again.
Base model
aifeifei798/QiMing-v1.0-14B