Intervention class
PRISM / PRISM-LITE / PRISM-PRO
Behavioral intervention applied to a named base model.
48 repositories: 37 public, 11 private. Search by upstream lab, model family, method, access, and date.
Ex0bit account snapshot · 14 July 2026.
Qwen
PRISM · NVFP4
mlx
Qwen/Qwen3.6-35B-A3B
Qwen
PRISM-PRO · PRISM-DQ · Speculative decoding
gguf
Qwen/Qwen3.6-27B
Qwen
PRISM · EAGLE-3
sglang
Qwen/Qwen3.6-27B
Qwen
PRISM-PRO
—
—
Qwen
PRISM · NVFP4
—
Qwen/Qwen3.6-35B-A3B
DeepSeek
PRISM-PRO · Tau
—
deepseek-ai/DeepSeek-V4-Pro
DeepSeek
PRISM-PRO
—
deepseek-ai/DeepSeek-V4-Pro
MiniMax
PRISM-DQ · SLURPY · Model merge · SLERP
mlx
Ex0bit/MiniMax-SLURPY
MiniMax
SLURPY · Model merge · SLERP
transformers
MiniMaxAI/MiniMax-M2.5
Google
PRISM-PRO · PRISM-DQ
llama.cpp
google/gemma-4-26B-A4B-it
Google
PRISM-PRO · PRISM-DQ
llama.cpp
google/gemma-4-31B-it
Qwen
PRISM · PRISM-DQ
llama.cpp
Qwen/Qwen3.5-0.8B
Qwen
PRISM-PRO
llama.cpp
Qwen/Qwen3.5-122B-A10B
Qwen
PRISM-LITE
llama.cpp
Qwen/Qwen3.5-122B-A10B
Qwen
PRISM-LITE
transformers
image-text-to-text
Qwen
PRISM-PRO
transformers
image-text-to-text
StepFun
PRISM-PRO · NVFP4
—
—
StepFun
PRISM-PRO
gguf
Ex0bit/Step-3.5-Flash-PRISM-PRO
StepFun
PRISM-PRO
transformers
stepfun-ai/Step-3.5-Flash
StepFun
PRISM
transformers
stepfun-ai/Step-3.5-Flash
MiniMax
PRISM-PRO
—
—
MiniMax
PRISM-PRO
—
—
MiniMax
PRISM-LITE
—
—
MiniMax
PRISM-LITE
transformers
MiniMaxAI/MiniMax-M2.5
MiniMax
PRISM-PRO
transformers
MiniMaxAI/MiniMax-M2.5
Moonshot AI
PRISM · REAP
transformers
moonshotai/Kimi-K2.5
Moonshot AI
PRISM
transformers
image-text-to-text
NVIDIA
PRISM · NVFP4
transformers
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
NVIDIA
PRISM
transformers
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
AI2
Abliteration · Refusal removal · SNR layer selection · Norm preserving · Orthogonalization
transformers
text-generation
AI2
Abliteration · Refusal removal · Triangular abliteration · Orthogonalization
transformers
allenai/Olmo-3-7B-Instruct
AI2
NVFP4
transformers
allenai/Olmo-3-7B-Instruct
Qwen
NVFP4
—
qingy2024/Qwen3-VLTO-32B-Instruct
Qwen
NVFP4
—
qingy2024/Qwen3-VLTO-32B-Instruct
PRISM is an intervention. PRISM-DQ is quantization. Runtime, hardware, access, and license are listed with the repository.
PRISM / PRISM-LITE / PRISM-PRO
Behavioral intervention applied to a named base model.
PRISM-DQ
Sensitivity-aware precision assignment by tensor class.
GGUF / safetensors · MLX / llama.cpp / SGLang · NVFP4 / FP8 / INT4
Artifact container, execution stack, and numerical representation.
Dense / MoE / multimodal · EAGLE-3 / native MTP / REAP
Base architecture and applied acceleration or compression method.
Qwen / Gemma / MiniMax / Kimi / StepFun / GLM / Nemotron / OLMo / Prime Intellect / LFM
Upstream weights and lineage.
Open / gated / member / one-time / quote-based
Download terms and commercial rights.
Weights, model card, deployment notes, and artifact license.
Browse public downloads ↗Request access on the model card or use a membership tier.
Compare membership tiers →Named artifact, deliverable, license, price, and checkout terms.
See listed release checkouts →Base weights, method, target hardware, evaluation, and operating rights.
Commission a model build ↗Model card, deliverable, license, price, and Ko-fi checkout.
Gemma 4 31B PRISM-PRO-DQ weights · GGUF / MLX
Kimi K2.5 PRISM weight tensors · 1T / 32B-active MoE
MiniMax M2.5 PRISM-PRO · multi-quant GGUF
Rights follow the artifact license and upstream base-model license.