← Back to search
This AI Model Has Native Text, Image, and Video Capabilities: Here's What You Should Know
Machine Learning Tech Brief By HackerNoon · 2026-10-10 · 16 min
Show full episode description
This story was originally published on HackerNoon at: https://hackernoon.com/this-ai-model-has-native-text-image-and-video-capabilities-heres-what-you-should-know . Qwen3.6-35B-A3B-Uncensored-Genesis-Hermes-GGUF is a locally runnable, text-to-text GGUF release with native text, image, and video capabilities. Check more stories related to machine-learning at: https://hackernoon.com/c/machine-learning . You can also check exclusive content about #artificial-intelligence , #ai-models , #ai-model-guide , #qwen , #text-to-text-gguf , #qwen3.6 , #qweun-guide , #qwen-best-use-cases , and more. This story was written by: @aimodels44 . Learn more about this writer by checking @aimodels44's about page, and for more stories, please visit hackernoon.com . The Qwen3.6-35B-A3B-Uncensored-Genesis-Hermes-GGUF model is a locally runnable GGUF model for text, image, and video tasks with 35 billion total parameters and 3 billion active per forward pass. It has a hybrid mixture-of-experts architecture and supports long context with a native 262K token limit, extendable to
✨ Episode Outline — click any point to jump to it in the episode
Problem solved
Explains what the Qwen3.6-35B-A3B Uncensored Genesis
Hermes GGUF model actually is, how to run it locally, and which of its claims are unverified.
Benefits
- Locally runnable GGUF with Hermes-derived agent behavior for tool-use conventions
- 262K native context (YaRN-extendable to 1M) for long documents and codebases
- Native text, image, and video modalities with llama.cpp, LM Studio, KoboldCpp support
- Documented sampling profiles for coding, general, and creative modes
- Apex Compact quantization guidance for 8GB and 12GB GPUs
Use cases
- Local coding and precise instruction following using the thinking-mode profile (temp 0.6, top_p 0.95, top_k 20, seed 42)
- Tool-oriented local agents using the JSON system-prompt pattern and Hermes function-calling conventions
- Long document and codebase analysis with at least 128K context to preserve thinking mode
- Uncensored creative writing and roleplay via the non-thinking creative profile (temp 0.7, top_p 0.8)
- Multimodal local experiments, requiring the separate vision mmproj file alongside the main GGUF
KPIs / results
- 35B total parameters, ~3B active per forward pass
- 262K native context, YaRN to 1M; minimum 128K recommended
- ~2,000 blocks transferred from two FFN expert tensors of a Hermes fine-tune
- Maintainer-reported zero refusals on a 465-prompt test; no independent benchmarks
Tools / build
- Qwen3.6-35B-A3B Uncensored Genesis Hermes GGUF
- Genesis Tensor Repair Process
- llama.cpp / LM Studio / KoboldCpp
- Apex and NVFP4 quantization
- Hermes function-calling dataset commands
This audio is presented by HackerNoon, where anyone can learn anything about any technology. This AI Model has Native Text, Image, and Video Capabilities. Here's What You Should Know, by AI Models 44. Overview, is a locally runnable, text-to-text GGUF release with native text, image, and video capabilities. Luffy the Fox built it from the uncensored HowHow CSQN3. 6-35 BA-3B base, applied the Genesis Tensor Repair Process, and transferred about 2,000 blocks from two FFN expert tensors in a Hermes fine-tuned to Adhermese agent behavior. It has 35 billion total parameters, about 3 billion active per forward pass, a 262k token native context window, and a hybrid MOE architecture combining gated DeltaNet linear attention with full softmax attention. The key practical point. This is a large GGUF model, not a lightweight 3B model. The active parameter count does not describe the full storage or memory footprint. It runs in LAMA, CPP, LM Studio, COBOLD CPP, and other GGUF compatible runtimes. The model card recommends NVFP4 or Apex quantization, and Apex Compact for 8GB and 12GB GPUs. The listed license is Apache 2.0, but check the terms of the upstream components before commercial deployment. Best use cases, local coding and precise instruction following. The model card provides a dedicated thinking mode sampling profile for coding and precise tasks, temperature 0.6, top underscore PO. 95, top underscore K20, min underscore PO. 01, seed 42, with presence Andre Pete penalties disabled. Its Hermes derived agent behavior and large context make it a candidate for code explanation, multi-file planning, and structured technical work. The README does not provide benchmark results, so test it against your own coding tasks before relying on it for production code. Tool-oriented assistant workflows. The Hermes agent transfer and the recommended function calling dataset commands make this release relevant to local agents that need to follow tool use conventions. The model card also gives a JSON-oriented system prompt pattern for schema-constrained responses. Treat these as prompting guidance, not proof of reliable tool execution. The README reports no function calling accuracy or structured output benchmark. Long document analysis. The native context is 262K tokens, and the model card says to keep at least 128K context to preserve thinking capabilities. That makes it suitable for experiments involving long reports, code bases, or document collections, subject to runtime memory and context cache costs. The model card says YARN can extend context to 1M, but it provides no quality or speed measurements at that length. Uncensored creative writing and role play. The base is explicitly described Austin-censored, and the maintainer reports zero refusals on a 465 prompt test for the base model. That result is a maintainer reported test, not an independent safety or quality evaluation. The card supplies a non-thinking creative profile, temperature 0, 7, top underscore PO, 8, top underscore K20, min underscore PO, 01, and optional creative system prompts. Multimodal local experiments. The architecture is described as natively multimodal for text, image, and video. For vision in GGUF, the README requires in-file alongside the main model file. The provided material does not specify supported image or video formats, resolution limits, or multimodal benchmark results, so validate those capabilities in the chosen runtime before building a workflow around them. Limitations. The model card's Genesis claims, including noise reduction, preservation of 99% of signal and learned gradient, and improved stability, are the maintainer's description of a custom tensor repair method. The supplied material does not include controlled comparisons, reproducible evaluation results, or independent validation. The model is not described as retrained or conventionally fine-tuned end-to-end. The maintainer says Genesis modifies GGUF tensor data and that HERMI's data was transferred from around 2,000 blocks in two FFN expert tensors. The uncensored base and the reported 0465th refusal result signal a safety tradeoff. Do not assume the model will refuse harmful requests or meet ad deployment safety requirements. The README also warns that context below 128 K may affect thinking capabilities, while the 1M yarn extension has no reported quality measurements. Vision requires a separate file. The main GGUF alone is not enough for that use. The listed license is Apache 2.0. The provided material does not clarify whether every upstream component and redistributed quantization carries identical terms, so review the relevant model and artifact licenses before commercial use. No specific newer model is identified as having superseded this release. How it compares? The supplied alternatives are other Genesis Hermes versions from the same maintainer. The provided information does not give controlled quality, speed, or cost comparisons between versions, so version numbers alone cannot establish which performs better. QUEN 3.6-35BA3B Uncensored Genesis Hermes V6GGUF Pick this release when you want the specific combination documented here. The HowHow CS Uncensored Base, Hermes Data Transfer from about 2000 blocks in two FFN Expert Tensors, and the described Genesis Repair. Pick V6 instead if its own model card documents a feature, quant, or runtime fit that your deployment needs. The supplied comparison data does not establish a quality or speed winner. QUEN 3.6-35BA3B Uncensored Genesis Hermes Final GGUF Choose this model when you need the exact release described in this guide and its stated settings and quant recommendations. Choose final when its artifact or card better matches your runtime or deployment requirements. No measured quality, latency, or memory comparison is provided here. QUEN 3.6-35BA3B Uncensored Genesis Hermes V10GGUF This release is the documented choice if you want the current model's stated 262K Native Context, Multimodal Support, and Apex NVFP4 Guidance. Consider V10 if its own card or files provide a better fit for your task. The supplied material does not establish that V10 is faster, more accurate, or cheaper to run. QUEN 3.6-35BA3B Uncensored Genesis Hermes V9GGUF Pick this release for the specific Genesis, Hermes construction and runtime instructions documented here. Pick V9 if you have validated it on your workload or need a version-specific artifact. No benchmark evidence in the supplied material supports a general quality or speed preference. QUEN 3.6-35BA3B Uncensored Genesis Hermes V7GGUF This release has explicit recommendations for Apex settings, context length, and multimodal use. V7 may suit you if its own files or card match your deployment better, but the provided information does not quantify differences in output quality, inference speed, or cost. Technical specifications, the model uses a mixture of experts architecture with 35B total parameters and about 3B active per forward pass. It has 256 experts, with 8 routed experts and 1 shared expert per token. Its 40 layers follow a repeating pattern of 3 DELTA NET MOE layers followed by 1 ATTENTION MOE layer. 10 repetitions in total. The ATTENTION design combines gated DELTA NET linear ATTENTION and full soft max ATTENTION in a 3 to 1 ratio. context, 262K tokens natively, extendable to 1M with yarn. The model card recommends at least 128K context to preserve thinking capabilities. Modalities, text, image, and video are listed as native capabilities. Vision support in GGUF requires and file alongside the main GGUF. Vocabulary and languages. 248K vocabulary, 201 languages. Model format and runtime. GGUF, compatible with LAMA. CPP, LM Studio, Kobold CPP, and other GGUF compatible runtimes. Use LAMA, CPP's flag for chat template handling. Quantization. Recommended options are NVFP4 and Apex. Apex Compact is recommended for GPUs with 8GB or 12GB of memory. The README links a quantization script with UNSLOTH profile support and points to a separate NVFP4 file. GGUF release. Apex settings. Set K and V cache quantization to F16, force MO weights for 40 layers onto CPU, set GPU offload to maximum, and set active experts to 8. Sampling profiles. Thinking mode Hermes agent and coding, precise tasks use temperature 0. 0. 6. TOP-UNDERSCORE PO. 95. TOP-UNDERSCORE K20, MIN-UNDERSCORE PO. 01. SEED 42, and disabled presence penalty. The Hermes agent profile uses REPEAT-UNDERSCORE PENALTY 1. 05. While coding, precise tasks disable REPEAT-PENALTY. Thinking mode General uses temperature 0. 95 with the same TOP-UNDERSCORE P, TOP-UNDERSCORE K, MIN-UNDERSCORE P, SEED, and disabled penalties. Non-thinking Creative General uses temperature 0. 7. TOP-UNDERSCORE PO. 8. TOP-UNDERSCORE K20, MIN-UNDERSCORE PO. 01. SEED 42, and disabled penalties. Genesis process, as described by the maintainer. The maintainer says the process scans tensors and adjusts head balance. Scans tensor blocks in chunks using three parameters and replaces zero chunks with a selected chunk. Then detects and reduces training noise with custom SVD using the Marchenko-Pastor law. The process excludes 1D tensors, biases, and norms. The maintainer says it runs on GGUF using Python and a free Google Colab Tesla T4 GPU, without retraining or fine-tuning the full model. Training and data. The README identifies the HowHow CS uncensored model as the base and a Hermes-Quinn 3. 5GGUF fine-tune as the source of transferred data. It reports around 2000 blocks from two FFN expert tensors. It gives no training step count, full dataset size, or training compute for the base model. Performance and memory. No measured tokens per second, latency, batch size guidance, or exact VRAM requirement is provided. The 8GB and 12GB figures refer to the maintainer's recommendation to use Apex Compact, not a guarantee that every context length or runtime configuration will fit. Model inputs and outputs. Inputs text. Prompts and conversation messages using the supplied chat template. In Lama. Cpp. Enable for proper template handling. Images. Supported as a native modality according to the model card. GGUF vision use requires the matching file. The README does not specify image formats or resolution limits. Video. Listed as a native modality. The README does not specify video container formats, frame handling, or limits. Context. Native context is 262k tokens. Yarn extension to 1M is stated, but the card recommends at least 128k to preserve thinking capabilities. Outputs text. Generated assistant responses, including coding, general, creative, and agent style responses. Structured text. The model card recommends a JSON system prompt with a supplied schema placeholder for agentic tasks. It does not report a structured output guarantee or validation rate. Multimodal responses. The model is described as natively multimodal, but the README does not specify output modalities or post-processing requirements beyond the requirement for vision input. Getting started. The README provides runtime and sampling settings, but no Python loading or inference example, and it does not identify a Python inference library or API. Our reliable copy-pasteable Python example cannot be derived from the supplied information. For a documented local setup, use a GGUF compatible runtime such as LAMA. CPP, LM Studio, or COBOLD CPP, in LAMA. CPP, pass, and include the file when using vision. Start with the recommended system prompt and keep context editor above 128k if you want to follow the maintainer's guidance for thinking mode. Frequently asked questions. Q. Can I use commercially? A. The listed license is Apache 2.0. The supplied material does not clarify the licensing of every upstream component or quantized artifact, so review those terms before commercial deployment. Q. What hardware or VRAM do I need to run it? A. The README gives no exact VRAM requirement. It recommends Apex Compact for 8GB and 12GB GPUs, and gives Apex offload settings, but those recommendations do not guarantee that every context length and runtime configuration will fit. Q. Is this model suitable for local tool calling agents? A. It includes Hermes agent behavior transferred from a Hermes fine-tune, and THE model card recommends a JSON system prompt and Hermes function calling commands. It provides no tool calling benchmark, so validate tool selection and argument formatting for your application. Q. What are the known failure modes or quality issues? A. The README does not report measured failure rates or benchmark results. It describes Genesis as a tensor repair method intended to reduce noise and improvestibility, but those claims are not accompanied by independent evaluation. The uncensored base also means you should not rely on it to refuse unsafe requests. Q. Can I fine-tune this model, and which framework supports it? A. The README does not document a fine-tuning workflow or name a fine-tuning framework. It links a quantization script with unsloth profile support, which is for quantization rather than evidence of a supported fine-tuning path. Q. What input format does the model expect? A. It uses GGUF compatible runtimes and a chat template. The model card recommends with LAMA. CPP. For vision, place the file alongside the main GGUF. The README does not specify image or video file formats. Q. How fast is inference, and what batch sizes are practical? A. The supplied material reports no inference speed, latency, or batch size measurements. Performance will depend on the quantization, runtime, hardware, context length, and offload settings. Q. Does it support long context and thinking mode? A. The stated native context is 262k tokens, with extension to 1M using YARN. The maintainer recommends at least 128k context to preserve thinking capabilities, but provides no quality measurements for the 1M extension. Feature image source. AI models. FYI this is a simplified guide to an AI model called QUEN 3. 6-35 BA-3B uncensored Genesis Hermes GGUF maintained by Luffy the Fox. If you like these kinds of analysis, join AI models. FYI or follow us on Twitter. Thank you for listening to this Hackernoon story, read by Artificial Intelligence. Visit HackerNoon.com to read, write, learn, and publish. and we find that young people are coming to one of the different places than a scout game. So, if you like this, make it