Open-Weight Models

Model · Text generation

granite-4.2-3b-MXFP4A16-GPTQ

David Reedy

An experimental MXFP4 quantization of ibm-granite/granite-4.2-3b, made with GPTQ calibration instead of the usual data-free rounding, and provided in two formats with identical weights. It's published as a data point, not as the recommended build. For most users, the Q4KM GGUF or GPTQ int4 build is the better choice. Format: OCP MX — E2M1 (FP4) weights, groupsize=32, one UE8M0 shared scale per group; weights only, activations stay 16-bit. Embeddings, norms and lmhead are unquantized. MXFP4 is usually produced data-free, by rounding weights straight to FP4 with no calibration data. llm-compressor ships an MXFP4A16 scheme, but every example we found pairs GPTQModifier with integer schemes…

Model · Text generation

granite-4.2-3b-Q4_K_M-imatrix-GGUF

David Reedy

A 4-bit GGUF of ibm-granite/granite-4.2-3b, quantized with an importance matrix built from a mixed prose and tool-calling calibration corpus. Of the three quantizations we made of this model, it's the closest to the original and the smallest. "Same top token" is llama.cpp's Same top p statistic: the share of positions where the quantized model's most likely next token matches the f16 model's. 1. Convert: converthftogguf.py (mainline llama.cpp), bf16 safetensors → f16 GGUF. Granite's scaling parameters (attentionscale, embeddingscale, residualscale, logitsscale) are supported natively. 2. Importance matrix: llama-imatrix over prose plus tool-calling conversations, rendered through Granite's…

Model · Text generation

granite-4.2-3b-W4A16-GPTQ-gs32

David Reedy

A 4-bit weight-only GPTQ quantization of ibm-granite/granite-4.2-3b, saved in compressed-tensors format for vLLM. On an AMD Radeon Pro V620 (RDNA2), vLLM served it at 65 tok/s single-sequence and ~1,000 tok/s aggregate at 16 concurrent sequences. llm-compressor GPTQModifier: - int4, symmetric, weight-only (activations stay 16-bit) - groupsize=32, actorder="weight" - lmhead left unquantized 512 samples, maxseqlength=512 The tuning mattered. With the default W4A16 preset (groupsize=128), the same pipeline scored 0.153 mean KLD. Moving to groupsize=32 with actorder="weight" cut that by about 28%, for roughly 0.2 GB more disk. For comparison, an AWQ build (W4A16 asymmetric, groupsize=128) from…

Model

granite-4.2-3b-oQ8e

Michael

This model was quantized using oQ (oMLX v0.7.0) mixed-precision quantization.

Model

granite-4.2-8b-oQ6e

Michael

This model was quantized using oQ (oMLX v0.7.0) mixed-precision quantization.

Model

granite-decisions

Adam Pippert

This is Adam Pippert's personal research project. This Hub repository publishes source code and a training recipe, not model weights or a tuned checkpoint. It cannot be loaded directly with frompretrained(). Source commit: 543345ea033370484ca226424afd73164d48ca35. Hub packaging adds this landing page and a copy of the project README; runtime code is unchanged. Fullcollar integration is deferred. This is not an IBM, Red Hat, or TypeSafe release. A local typed-decision runtime and reproducible training recipe around The code supports Choice distributions, ordered Score rubrics, and Boolean Noul probabilities. It downloads official IBM weights; no new pretrained model or production-quality…

Model

granite-guardian-4.1-8b-smoothquant-w8a8

Marcus Nogueira

SmoothQuant + GPTQ W8A8 INT8 checkpoint of ibm-granite/granite-guardian-4.1-8b at immutable revision ab01ccca5dcfb80246369a086a4a87a29198f5af. - Smoothing strength (alpha): 0.5 The resolved base revision, full quantization configuration, package versions, architecture mapping inventory, tensor inventory, and file hashes are included in the checkpoint metadata. vLLM smoke-test evidence is added before publication. This checkpoint passed four fixed English/Portuguese safe/unsafe smoke cases in BF16 with vLLM 0.31.0 on an NVIDIA A40. vLLM loaded the serialized compressed-tensors W8A8 configuration directly; no on-the-fly quantization was used. BF16 is the selected runtime compute dtype for the…

Model

granite-jev-1b-sampling

OVRLab

A runnable inference procedure, using unchanged IBM Granite 4.0–1B weights and TypeSafe's hosted Jev verifier. This release implements the current There are no newly trained weights, adapters, internal routing changes or Jev weights in this package. It is a custom serial Python runner, not an AutoModel checkpoint, hosted Hugging Face chat widget or vLLM plugin. The earlier trained-proposal artifact is a different preserved experiment. 1. Frozen Granite generates one greedy answer. 2. The same weights generate one independent sample (temperature 0.7, top-p 0.95, top-k 50), with a separate cache and a reproducible input-specific seed. 3. Exactly equal texts keep the native answer without…

Model · Speech recognition

granite-speech-4.1-2b-nar-glade

Josh Leverette

Glade is a native Swift library for on-device speech inference on Apple silicon. This repository supplies converted model assets and configuration for Glade's model-specific host runtime. Glade's runtime source is currently private. A community conversion of IBM Granite Speech 4.1 2B NAR for Glade, using Apple's Core AI APIs and optimized for Neural Engine execution on Apple silicon. The model combines a CTC acoustic encoder with one bidirectional, audio-conditioned transcript-editing pass. This is a W8A16 export: most convolution/projection weights are INT8 and activations are FP16. Conformer convolution output projections, normalization parameters, and the embedding table remain FP16.…

Model · Speech recognition

granite-speech-5.0-470m-turboctc-glade

Josh Leverette

Glade is a native Swift library for on-device speech inference on Apple silicon. This repository supplies converted model assets and configuration for Glade's model-specific host runtime. Glade's runtime source is currently private. Glade source asset for IBM's English CTC speech recognizer, pinned to The upstream model and these converted weights use Apache License 2.0. Requires physical Apple silicon with macOS 27 or iOS 27 and Glade's host runtime. This is a source.aimodel, which Core AI specializes on the device. It is not a precompiled.aimodelc or a standalone transcription application. One static entry point, frames3072, handles up to 3,072 feature positions at approximately 50…

Model

granite41-8b-squad-ratio-0.50-seed-42-r4

Alina Hancharova

This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

Model

granite41-8b-squad-ratio-0.60-seed-42

Alina Hancharova

This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

A multi-output Random Forest regressor for predicting residential property energy consumption and valuation, built for the Lloyds Green Home Score ESG intelligence platform. This model predicts three targets simultaneously from 13 property features: - Synthetic simulator consistency only, not real-world accuracy - Current value model validation is synthetic, not historical UK transaction validation - Value ranges are sensitivity bands, not calibrated confidence intervals Part of the Lloyds Green Home Score platform for ESG risk assessment, retrofit recommendation, and green lending analytics. Not intended for direct property valuation or energy certificate prediction. - multioutputrf.joblib…

This is a merged QLoRA derivative of Qwen/Qwen3-4B-Instruct-2507, published for Headroom, developed under the working name Greenlight. The original model identifier is retained for reproducibility. Training used 336 synthetic examples and 300 configured optimizer steps through Veri. Dataset SHA-256: ed827924d7ba4118fa15c5ef71fb211c557c3390d87384f21af4eb86ae59d051. Training job: 42d582a35d2f. Rank 8, 4-bit loading, learning rate 0.0001, sequence limit 2048, micro-batch 1, gradient accumulation 2. Prompt profile: reference-contract-v5. This model proposes bounded structured payment terms. Consult the repository’s matched frozen base/tuned evaluation before making any quality claim. The corpus…

Model · Robotics

groot-n17-droid-pour-water-10ep

Sang Nguyen

GR00T N1.7 is an open, cross-embodiment foundation model from NVIDIA for generalized humanoid robot reasoning and skills. It uses a Cosmos-Reason2/Qwen3-VL backbone and a flow-matching action transformer to predict actions conditioned on vision, language, and proprioception. This policy has been trained and pushed to the Hub using LeRobot. Learn how to train and run it in the LeRobot groot guide, or browse the full documentation. The policy consumes these observation features and produces these action features. Inputs Outputs New to LeRobot? These guides cover the full workflow: - Install LeRobot — set up the lerobot package. - Hardware setup — assemble, wire, and calibrate your robot and…

GR00T-N1.5-3B fine-tuned with LeRobot (0.4.1) on the 379 LIBERO-10 (LIBERO-Long) demonstrations, on a single RTX 4090. This is the final 60,000-step checkpoint. It is a color-jitter ablation of an RGB-vs-grayscale study. The recipe is identical to except that training-time color jitter is turned off. It reaches 66% on LIBERO-10 under default lighting. It is much less robust to lighting changes than the jittered RGB and grayscale models. zhoumiaosen/libero10rgb: the 379 LIBERO-10 episodes of openvla/modifiedliberorlds, converted to LeRobot v3.0 at 10 fps. That is 101,469 frames. observation.state (8-D: eef position, axis-angle, 2 gripper joints). Output: 7-D action (eef delta + gripper).…

Model

grounded-language-base57

Anna Szymanski

Reading notes and an experiment sketch for Grounded Language. The repository emphasizes what still needs to be tested instead of manufacturing scores or release claims. - the scope of the research question and likely confounders - a proposed comparison with matched baselines - concrete evaluation context such as RefCOCO, Flickr30k, and Visual Genome - reproducibility checks, failure modes, and open questions - topic-relevant references Start with review.md for the full note. Sections labeled as plans or hypotheses should not be interpreted as experimental results. If results are added later, they should include dataset versions, commands, seeds, hardware, and raw logs. The note is…