---
language: en
license: mit
library_name: mlx
base_model: ShaunGves/FastContext-1.0-4B-SFT
tags:
- mlx
- code
- agent
- repository-exploration
pipeline_tag: text-generation
---

# FastContext-1.0-4B-SFT — MLX 4-bit

MLX 4-bit quantization (4.501 bits per weight, ~2.1 GB) of
[ShaunGves/FastContext-1.0-4B-SFT](https://huggingface.co/ShaunGves/FastContext-1.0-4B-SFT), a surviving
mirror of Microsoft's FastContext repository-exploration subagent
([arXiv:2606.14066](https://arxiv.org/abs/2606.14066)), removed from the official listings on 2026-06-30.
Base model: Qwen3-4B-Instruct-2507. Converted with mlx-lm 0.29.1 (default group size 64).

FastContext is a small explorer model for coding agents: given a natural-language query about a repository,
it explores with read-only tools (Read / Glob / Grep, called in parallel) and returns a compact
`<final_answer>` block of `file:line-range` citations, keeping broad exploration out of the main agent's
context window.

## ⚠️ Quality note (tested)

In end-to-end tests with the [fastcontext CLI](https://github.com/manjunathshiva/fastcontext) on an M4 Mac,
this 4-bit quant showed **noticeably degraded path grounding**: it repeatedly assumed wrong repository root
paths, failed to recover from tool errors, and hallucinated citations to nonexistent files — on the same
queries the [8-bit quant](https://huggingface.co/mlx-community/FastContext-1.0-4B-SFT-8bit) answered
accurately. It is published for memory-constrained setups and further experimentation;
**prefer the 8-bit quant** (~4 GB) if your machine allows. Run with `temperature ≤ 0.6` and cap
`max_tokens` (missed stop tokens otherwise generate for minutes).

## Use with LM Studio

Search for `FastContext-1.0-4B-SFT-4bit` in LM Studio (MLX runtime), or:

```bash
lms get mlx-community/FastContext-1.0-4B-SFT-4bit
```

## Use with mlx-lm

```bash
uv tool install "mlx-lm==0.29.1" --with "transformers<5" --with "mlx<0.31"
mlx_lm.server --model mlx-community/FastContext-1.0-4B-SFT-4bit --port 8080
```

## Use with the fastcontext CLI

Install from the [preserved mirror](https://github.com/manjunathshiva/fastcontext) (includes fixes for
local OpenAI-compatible servers), then from the repo you want to explore:

```bash
export BASE_URL="http://localhost:8080/v1"   # or http://localhost:1234/v1 for LM Studio
export MODEL="mlx-community/FastContext-1.0-4B-SFT-4bit"
export API_KEY="local"
export TEMPERATURE=0.6
export MAX_TOKENS=4000

fastcontext -q "Where is the retry logic for failed API calls?" --citation
```