OpenZero Qwen3-1.7B Agentic is a compact local GGUF model fine-tuned for coding, research, debugging and tool-use workflows. Choose the verified Q4_K_M, Q8_0 or F16 build and run it with llama.cpp-compatible tooling.

What is included

File	Purpose
`OpenZero-Qwen3-1.7B-Agentic-Q4_K_M.gguf`	Recommended balance of size and quality
`OpenZero-Qwen3-1.7B-Agentic-Q8_0.gguf`	Higher fidelity, larger download
`OpenZero-Qwen3-1.7B-Agentic-F16.gguf`	Reference full-precision GGUF

Standalone model: yes
Separate adapter required: no
Separate base model required: no
Base architecture: Qwen/Qwen3-1.7B
Fine-tuning: 2,606 training examples; 137 held-out evaluation examples
Final held-out loss: 2.270656
GGUF conversion and CPU load test: passed for all three files with llama.cpp b10333

Run with llama.cpp

hf download shafire/OpenZero-Qwen3-1.7B-Agentic-GGUF OpenZero-Qwen3-1.7B-Agentic-Q4_K_M.gguf --local-dir .

llama-cli -m OpenZero-Qwen3-1.7B-Agentic-Q4_K_M.gguf --jinja -c 8192 -t 8 --temp 0.6 --top-p 0.95

For a local OpenAI-compatible endpoint:

llama-server -m OpenZero-Qwen3-1.7B-Agentic-Q4_K_M.gguf --jinja -c 8192 -t 8 --host 127.0.0.1 --port 8080

Positioning

Built for private local inference, practical code work, evidence-aware research and agent runtimes. Tool execution belongs to the surrounding runtime; validate outputs before acting on them.

Verified release

Training adapter: QLoRA specialist run, one epoch
Train loss: 2.949538
Held-out evaluation loss: 2.270656
Q4_K_M — 1,107,408,576 bytes — SHA-256 1d43348dc10a4b97ec733cc435e398393cd235f7f6088b0cced2382ed8c9b1b7
Q8_0 — 1,834,426,048 bytes — SHA-256 b95cedcf23e5698fd2e6368caa18e8f38ec953de5a31f449f0f18bd9993cfb0a
F16 — 3,447,348,928 bytes — SHA-256 4d0de653af5b248dfbcb0a75afcc9a8d30c261e34a98efa1208bf1e51498fc70

Provenance and reproducibility

The V10 LoRA adapter was trained for one epoch on 2,606 OpenZero instruction examples. A separate 137-row held-out set was used only for final evaluation. The adapter was merged into Qwen/Qwen3-1.7B revision 70d244cc86ccca08cf5af4e1e306ecf908b1ad5e, converted with llama.cpp b10333 / commit 08659901c43b51de735740f1cf61bb82fbe0c4e4, and Q8_0 and Q4_K_M were independently quantized from that F16 source. Every file passed a bounded one-shot CPU text load test (-c 128 -n 1 -ngl 0 --no-conversation --single-turn --simple-io --no-warmup) and its SHA-256 was compared with the remote Hugging Face LFS/Xet object.

This is a practical specialist fine-tune, not a claim of superiority over the base model. Tool calls are text emitted for an agent runtime to validate and execute; the GGUF does not itself access a shell, browser, network or private data.

This model is an independent fine-tune based on Qwen and is not affiliated with or endorsed by Qwen. The Qwen base is Apache-2.0. OpenZero training materials and resulting community release are subject to the OpenZero Community Source terms; do not describe this release as OSI-approved open source. Review both upstream and OpenZero terms before redistribution or commercial use.