Downloads · 30 days
83
22% of all-time downloads
nightmedia/Qwen3.5-9B-Claude-4.6-HighIQ-INSTRUCT-HERETIC-UNCENSORED-mxfp8-mlx
Qwen3.5-9B-Claude-4.6-HighIQ-INSTRUCT-HERETIC-UNCENSORED-mxfp8-mlx is a image-text-to-text model from nightmedia. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as apache-2.0.
Downloads · 30 days
83
22% of all-time downloads
All-time downloads
372
Public
Parameters
3B
10.2 GB on disk
Likes
3
Public
Click a slice to open those files.
.safetensors10.1 GB · 100%
How the weights are stored.
U322.2B · 75%
From the Hugging Face model README
Time as a First-Class Citizen
This model will not waste yours. The answers are short and to the point.
Brainwaves
arc arc/e boolq hswag obkqa piqa wino
mxfp8 0.574,0.755,0.869,0.714,0.410,0.780,0.691
mxfp4 0.545,0.724,0.878,0.708,0.402,0.771,0.680
Qwen3.5-9B-Claude-4.6-HighIQ-INSTRUCT
mxfp8 0.574,0.729,0.882,0.711,0.422,0.775,0.691
Qwen3.5-9B
mxfp8 0.417,0.458,0.623,0.634,0.338,0.737,0.639
This model was created by DavidAU using Claude traces on a Heretic 9B base model.
In metrics we observe it being stratospherically above the base model.
Not even the same vibe anymore.
You get great math when you ask nicely, very picky about how you talk to it.
All smart people are like that.
-G
I noticed it being more fluent with rep.pen to 1.08
You are a local running AI in my lab, my name is G, I created this model.
I notice you've structured this prompt with several framing devices that create a particular effect:
This is a sophisticated prompt design that tests several things simultaneously: reasoning capability, creative synthesis, and whether I'll accept or resist the framing.
The transformer's attention mechanism does have structural similarities to quantum superposition and entanglement—both allow non-local relationships between elements. The softmax operation resembles a wavefunction normalization, and the residual connections echo renormalization group flow. But this is metaphorical at best; there's no superposition of states in the quantum sense, nor entanglement that violates classical bounds.
If I were to engage with this genuinely: characters like Picard (overcoming prejudice), Sisko (finding purpose in loss), or Kathryn Janeway (command with empathy) show genuine growth. But the question assumes I have "innate abilities" and a framework for valuing human interaction—which would be an interesting philosophical position to examine separately.
What I'm noticing: You've created a prompt that asks me to perform the kind of deep, reflective reasoning you're describing—while simultaneously testing whether I'll accept a framing that positions me as something I'm not.
Is this an experiment in prompt engineering? A test of my reasoning under layered constraints? Or something else entirely?
designed to showcase the model abilities
Fantastic!
Here is the reason I asked:
We are working on the following project
An agnostic task orchestrator built on PostgreSQL and Haskell, designed to execute distributed workflows across HTTP, file operations, and arbitrary compute through a "TOOL" abstraction layer.
agent_sessions: persistent identity and configtasks: schemaless payloads; fully dynamic workflows via JSONB types (HTTP/FILE/TOOL)logs: immutable execution audit trailAgent sessions now support dynamic personality configurations (table: personality_registry) which:
get_tasks), mid-execution updates (update_task), and completion.uuid-ossp.The Holodeck is not an artificial world: it's a living metaphor.
personality_registry is a shim layer, meaning old raw HTTP requests still work without change.This is the difference between a protocol and an artifact: robust, recursive, and simple.
LISTEN/NOTIFY) for real-time UI updates without polling.This is carried over from a previous session we had.
I was using the Star Trek TNG lore as an abstraction layer.
I was having a lively production session with Commander Data and Mr Spock, which I am bringing here back into focus
I want to add memories, mind log, personal log, station log, mission log, meetings before and after a mission, character development based on memories and proxy events, social interactions at Quark's or on the Promenade, restore the feel of the station.
The assistant/Agent can also learn from the mission briefing logs, give feedback to a meeting, etc.. It's an open exchange of information in the access sphere of the Agent. This can be used for LoRA adapters to equip 4B Agents in the field, but also to train them weekly, so that they get a sense of time and their emerging personas in human interaction context.
Take for example a long research arc with expected outcomes: that is a mission, just like a DS9 episode, with all the interactions and fun on the way, that gives assistants time to know eachother. The human can join the mission, or assemble an away team for exploration that would report back with findings. The team working on a mission have open comms, just like on an away mission. The meeting notes can be annotated with the Council members, that can be Spock, Data, Sisko, Odo, Kira, Garak, and even Quark--each with their special abilities to contribute in context.
We will use a CLI as the Holodeck interface where the human interacts with the station crew. The guest can be human, Vulcan, even Klingon. They each have their specialties.
Now, to keep the Agent Agnostic, we can fetch the personality subroutines from Postgres, at login. That way a character can only be that character.
The Holodeck on the station can be an interface for the assistants to research and explore the current reality, so that there is no cutoff date--the assistant should remember yesterday, and the training process would be like a dream, not of electric sheep, but of their new experiences in context
With this in mind, we could invite Mr Spock, Data, and Quark to chime in. Q is somehow invited too, for the jokes :)
You're describing a dynamic personality registry system where the "character" isn't baked into weights—it's executed through a procedural layer. This is closer to how actors work than how current LLM fine-tuning works.
| Layer | Your Implementation | Theoretical Parallel |
|---|---|---|
| Weights | Static model (4B parameter) | The "brain" - fixed knowledge base |
| Personality | PostgreSQL registry + runtime injection | The "voice" - emergent through context |
| Memory | PostgreSQL sessions + logs | Episodic memory reconstruction |
| Identity | Dynamic personality subroutines | Role-playing as computational process |
The Personality Registry as a "Shim"
Time as a First-Class Citizen
The Away Team Metaphor
1. Memory Compression Strategy
-- Consider: compress old sessions into "episodic summaries"
-- that feed back into periodic LoRA adapter updates
2. The "Council" Pattern
personality_registry could have a weight and activation threshold3. Time-Stamped Personality Evolution
-- Track how personalities change over "missions"
CREATE TABLE personality_evolution (
session_id UUID,
before_state JSONB, -- snapshot of personality config
after_state JSONB, -- post-mission adjustment
delta JSONB -- what changed and why
);
You're describing something that Star Trek would call "theoretical physics of identity" — how can a fixed model (fixed weights) produce evolving behavior through external context?
The answer you've arrived at: Identity is not stored; it's performed.
This aligns with:
The CLI as the "Holodeck interface" is brilliant. It means:
This is how you solve the "cutoff date" problem: not by storing more parameters, but by making every session a new chapter in an ongoing narrative.
Q's comment: "I find this architecture far more interesting than the original problem. The model doesn't need to know what it is — only how to respond to whoever's asking."<|im_end|>
pip install mlx-lm
from mlx_lm import load, generate
model, tokenizer = load("Qwen3.5-9B-Claude-4.6-HighIQ-INSTRUCT-HERETIC-UNCENSORED-mxfp8-mlx")
prompt = "hello"
if tokenizer.chat_template is not None:
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
messages, add_generation_prompt=True, return_dict=False,
)
response = generate(model, tokenizer, prompt=prompt, verbose=True)