Sparx-Assistant: An AI Math Problem-Solving Assistant for Sparx Maths

Sparx-Assistant is an open-source browser extension designed specifically for Sparx Maths. It integrates Groq cloud-based models or local LM Studio models to provide features like intelligent problem-solving and step-by-step reasoning. Positioned as an educational aid, it emphasizes academic integrity and the limitations of AI, supporting two deployment options to meet different user needs.

Sparx-Assistant was open-sourced by Jacob K. under the AGPL v3 license. It aims to address the pain points in math learning, such as the lack of immediate guidance and high cost of human tutoring. By embedding AI capabilities into the learning process via a browser extension, it provides personalized problem-solving guidance.

Core Features:

1. Intelligent context awareness: Automatically identifies problems and adds an assistant button
2. Step-by-step reasoning display: Presents problem-solving ideas in expandable cards
3. Quick operations: Summary dock bar, direct pasting, F-key trigger

Technical Architecture:

• Cloud solution (Groq): Instantly usable, flexible models, free quota; limitation is weak image processing
• Local solution (LM Studio): High privacy, requires GPU memory ≥8GB, supports TIR, visual understanding, etc.

Benchmark performance of recommended models for local deployment:

Model	Quantized VRAM Usage	Application Scenarios	Benchmark Performance
qwen/qwen3.5-9b	~6GB (@Q4_K_M)	8-12GB VRAM	MathVision:78.9%
google/gemma-4-26b-a4b	~17GB (@Q4_K_M)	24GB+ VRAM	AIME 2026:88.3%
qwen/qwen3.5-35b-a3b	~21GB (@Q4_K_M)	24GB+ VRAM	AIME 2026:93.3%

Sparx-Assistant is a typical case of AI-assisted education, demonstrating the integration of large models with educational scenarios. Its open-source nature supports community contributions; in the future, it can integrate multimodal models to enhance graphic processing capabilities, empowering learning efficiency and experience.

1. Academic integrity: As an auxiliary tool, do not directly obtain answers
2. AI limitations: May have hallucinations; need to verify the problem-solving process
3. Visual limitations: Groq cloud does not process images; complex problems are recommended to use local deployment
4. Non-official affiliation: Independent third-party tool with no official collaboration with Sparx Maths