digitranslab/Megamind-v2-VL-high-gguf

Megamind-v2-VL: Multimodal Agent for Long-Horizon Tasks

Overview

Megamind-v2-VL is an 8B-parameter vision–language model for long-horizon, multi-step tasks in real software environments (e.g., browsers and desktop apps). It combines language reasoning with visual perception to follow complex instructions, maintain intermediate state, and recover from minor execution errors.

We recognize the importance of long-horizon execution for real-world tasks, where small per-step gains compound into much longer successful chains—so Megamind-v2-VL is built for stable, many-step execution. For evaluation, we use The Illusion of Diminishing Returns: Measuring Long-Horizon Execution in LLMs, which measures execution length. This benchmark aligns with public consensus on what makes a strong coding model—steady, low-drift step execution—suggesting that robust long-horizon ability closely tracks better user experience.

Variants

• Megamind-v2-VL-low — efficiency-oriented, lower latency
• Megamind-v2-VL-med — balanced latency/quality
• Megamind-v2-VL-high — deeper reasoning; higher think time

Intended Use

Tasks where the plan and/or knowledge can be provided up front, and success hinges on stable, many-step execution with minimal drift:

• Agentic automation & UI control: Stepwise operation in browsers/desktop apps with screenshot grounding and tool calls (e.g., BrowserMCP).

Model Performance

Compared with its base (Qwen-3-VL-8B-Thinking), Megamind-v2-VL shows no degradation on standard text-only and vision tasks—and is slightly better on several—while delivering stronger long-horizon execution on the Illusion of Diminishing Returns benchmark.

Local Deployment

Integration with Megamind

Megamind-v2-VL is optimized for direct integration with the Megamind. Simply select the model from the Megamind interface for immediate access to its full capabilities.

Local Deployment

Using vLLM:

vllm serve digitranslab/Megamind-v2-VL-high \
    --host 0.0.0.0 \
    --port 1234 \
    --enable-auto-tool-choice \
    --tool-call-parser hermes \
    --reasoning-parser qwen3 
    

Using llama.cpp:

llama-server --model Megamind-v2-VL-high-Q8_0.gguf \
    --vision-model-path mmproj-Megamind-v2-VL-high.gguf \
    --host 0.0.0.0 \
    --port 1234 \
    --jinja \
    --no-context-shift

Recommended Parameters

For optimal performance in agentic and general tasks, we recommend the following inference parameters:

temperature: 1.0
top_p: 0.95
top_k: 20
repetition_penalty: 1.0
presence_penalty: 1.5

🤝 Community & Support

• Discussions: Hugging Face Community
• Megamind: Learn more about the Megamind at megamind.ai

📄 Citation

Updated Soon