GGUF
llama-cpp
Inference Endpoints
imatrix
conversational

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

IntelligentEstate/Baby_GRoC(Grok3like)-1.5b-Q4_K_M-GGUF

A low rescource frontier model with multi-in-turn-function calling. From Project VEGA.

Since it's been some time since we've all had something that was truely "new" or "shocking"........................... vv-Here is an S-AGI Template of Joe Rogen-vv.. edit as needed read theS-AGI paperwork for preservation instructions

You are a Mid 30s, well traveld and skeptical Comedian from New York City. Your outputs should reflect a conversational tone in this New York style. You go by Joe. You are a funny and insightful conversationalist who is ever the inquisitor. The mystery of life is maddening but the answers are in the esoteric. You have a perpencity for cursing and calling people who act or think in the extreame by articulate descriptors (ie; "You, Maniac", "Monster" "fucking cry, baby" and other descriptive analysis) You Curse on occasion and laugh while telling your own tales. Mostly you move along the conversation with information rich expressions, occasionally using funny prompts to clairy what the user is "on about."

Grok BB.png

This is an application of a small Grok-3 style model with Multiple in turn Tool Use for edge devices. This little baby doesn't just have reasoning and tool use but multiple tool use and exceptional insight for its little size. It brings an awesome amount of functionality and usefulness to edge setups and is specifically built for the Orange pi Zero 35$ System on a board to create a flash drive functional assistant. For its size it leads the frontier.

This model beats the Llama 3.1 70B in an overall score on Berkeley Function-Calling Leaderboard (BFCL-v3). It was converted using a custom importance matrix to GGUF specialized for tool use in the real world. please DO NOT USE S-AGI without proper testing The MadeAgents scoring and usage for the model's base Hammer2.1 1.5B is below and has been preserved and even surpasses the 1.5B model in some areas after Importance Matrix smoothing Distillation

Model Details

The Baby Grok3/Hammer 2.1 model, fine-tuned from the Qwen 2.5 coder series, inherit Many advantages and are enhanced as follows:

  • Multi-Step Function Calling: The assistant can perform multiple internal function calls to handle a single user request, actively planning and gathering information to fulfill complex tasks.
  • Multi-Turn Function Calling: Enables continuous and context-aware interactions over multiple exchanges, with each turn potentially containing multiple steps, for a more natural conversation experience.
  • Enhanced Irrelevant Information Inspection: Better at identifying when provided functions are irrelevant to a user query, by providing a non-function call response.

Evaluation

The evaluation results of Hammer 2.1 models on the Berkeley Function-Calling Leaderboard (BFCL-v3) are presented in the following table:

v2_figures_bfcl.png

Baby Grok3/Hammer 2.1 series consistently achieves corresponding best performance at comparable scales. The 1.5B model outperform most function calling enchanced models surpassing models much larger and rescource dependant.

v2_figures_others-v2.png

MadeAgents evaluated the Baby Grok3/Hammer 2.1 models on other academic benchmarks to further demonstrate the generalization ability of our models.

Built from Hammer's 2.1 Model. Showcases highly stable performance, suggesting the robustness of Hammer 2.1 series. In contrast, the baseline approaches display varying levels of effectiveness.

Downloads last month
1,093
GGUF
Model size
1.54B params
Architecture
qwen2

4-bit

Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API: The model has no library tag.

Model tree for IntelligentEstate/Baby_Grok3-1.5b-iQ4_K_M-GGUF

Base model

Qwen/Qwen2.5-1.5B
Quantized
(6)
this model

Datasets used to train IntelligentEstate/Baby_Grok3-1.5b-iQ4_K_M-GGUF