---

license: apache-2.0

pipeline_tag: text-generation

tags:

• gguf
• llama.cpp
• pebble
• mamba2
• hybrid
• small-language-model
• chat

base_model: basically-ai/Pebble-10M-Chat

---

Pebble-10M-Chat-GGUF

GGUF conversions of basically-ai/Pebble-10M-Chat (Apache 2.0).

IMPORTANT: patched llama.cpp required

Pebble uses a custom hybrid Mamba2 + attention architecture. These GGUFs carry

general.architecture = "pebble", which upstream llama.cpp refuses to load.

Everything needed to run them lives in the support repo:

rootendpoint/basicallyai_llama.cpp_support

• llama.cpp-pebble.patch - adds the pebble architecture to llama.cpp

(applies cleanly against upstream commit 0eadefe)

• basicallyai_to_gguf.py - standalone converter (numpy + safetensors only)
• numpy_reference.py - independent reference implementation used to verify correctness

Apply the patch, rebuild llama.cpp, then:

llama-cli -m pebble-10m-chat-f16.gguf

Chat template

The lowercase user: / assistant: label format is embedded in the GGUF.

It was reverse-engineered by A/B testing and confirmed by the basically-ai

team as the exact SFT format:

user: <message>
assistant: <response>

llama-cli and llama-server pick it up automatically.

llama-server -m pebble-10m-chat-f16.gguf --host 127.0.0.1 --port 8080
# web UI at http://127.0.0.1:8080

Files

| Quant | Size | Type |

|-------|------|------|

| f16 | 20.7 MB | F16 |

| q8_0 | 11.1 MB | mostly Q8_0 |

| q4_k_m | 7.5 MB | mostly Q4_K_M |

Verification

Outputs were cross-checked token-by-token against an independent pure-numpy

reference implementation over multiple prompts (CPU and CUDA backends, and

quantized KV cache). Identical greedy sequences up to genuine argmax ties.

Model

• 10M parameters, hidden 384, 8 layers (6 Mamba2 + 2 attention), ctx 512, vocab 2048
• SFT on 250M tokens of smol-smoltalk
• A research-scale model: expect toy-level output quality.

CPU support (no GPU required)

Pure-CPU support for these models (no mamba-ssm, no CUDA) lives in the

basicallyai_cpu_support repository:

https://github.com/rootendpoint/basicallyai_cpu_support

It runs the original HF checkpoints in pure PyTorch on plain CPU

(~118 tok/s for 10M, ~65 tok/s for 25M on a Ryzen 5 2600X), verified

token-identical against this GGUF pipeline and an independent numpy oracle.