codegeist/codegeist-llm

1---2base_model: Qwen/Qwen3-1.7B3pipeline_tag: text-generation4inference: false5language:6- en7license: other8license_name: apache-2.0-and-0bsd9license_link: https://huggingface.co/codegeist/codegeist-llm/blob/main/THIRD_PARTY_NOTICES.md10tags:11- gguf12- q4_k_m13- docker-model-runner14- peft15- lora16- codegeist-training17---18 19# Codegeist LLM Qwen3-1.7B Artifacts20 21This experimental release adds one complete merged Q4_K_M GGUF for Docker Model22Runner while retaining the original first-stage PEFT adapter.23 24```text25User: What is Codegeist?26Assistant: Codegeist is a coding agent created by René Schmidt.27```28 29## GGUF Identity30 31| Field | Value |32| --- | --- |33| Release | `v0.3.0-alpha.3` |34| File | `gguf/codegeist-llm-Q4_K_M.gguf` |35| Size | `1107408672` bytes |36| SHA-256 | `be7824de2fc34955d640e30e41e92dd66206e86ab7fe027084015a9b7da44fce` |37| Quantization | `Q4_K_M` without an importance matrix |38| Default generation mode | Non-thinking; explicit thinking remains available |39| Base revision | `70d244cc86ccca08cf5af4e1e306ecf908b1ad5e` |40| Adapter revision | `a9504a0ee1150ea05f88ff725758404fcb604a32` |41| llama.cpp | `08659901c43b51de735740f1cf61bb82fbe0c4e4` |42 43## Docker Model Runner44 45The short command selects Q4_K_M from the mutable Hub `main` revision:46 47```bash48docker model run hf.co/codegeist/codegeist-llm:Q4_K_M   "What is Codegeist?"49```50 51The GGUF defaults to non-thinking even when a runtime enables Qwen thinking by52default. Add `/think` to a prompt to opt in explicitly.53 54Security-sensitive consumers must instead download this exact file from the55recorded release commit, verify the SHA-256 above, and package the verified local56file with `docker model package --gguf`.57 58## Scope And Limits59 60The first adapter and this merged GGUF establish model identity only. They do not establish coding ability, reasoning, generalization, tool use, safety,61Vulkan deployment, complete GPU offload, Codegeist OS integration, or62production release quality. This unsigned alpha artifact is Docker Model Runner63interoperability evidence, not the T001 release model.64 65## Licenses And Provenance66 67The merged GGUF includes Qwen3-1.7B weights distributed by Qwen under68Apache-2.0. The Codegeist-authored adapter, record, and project documentation use690BSD. See `THIRD_PARTY_NOTICES.md`, `gguf/QWEN3-1.7B-LICENSE.txt`, and the JSON70records under `gguf/` for exact revisions and transformation evidence.71