Interfaze

logo

Beta

pricing

help

docs

blog

sign in

Qwopus3.6 27B V2 MTP GGUF

Qwopus3.6 27B V2 MTP GGUF by Jackrong, a text-generation model with multimodal capabilities. Understand and compare multimodal features, benchmarks, and capabilities.

Comparison

FeatureQwopus3.6 27B V2 MTP GGUFInterfaze
Input Modalities

text, image

image, text, audio, video, document

Native OCRNoYes
Long Document ProcessingNoYes
Language Support

162 partial

162+

Native Speech-to-TextNoYes
Native Object DetectionNoYes
Guardrail ControlsYesYes
Context Input Size

1M

1M

Tool CallingYes

Tool calling supported + built in browser, code execution and web search

Scaling

FeatureQwopus3.6 27B V2 MTP GGUFInterfaze
Scaling

Self-hosted/Provider-hosted with quantization

Unlimited

View model card on Hugging Face

💡 1. Base Model, Training Library & Cooperation

[!WARNING] Community Release Notice: Qwopus3.6-27B-v2-MTP is an experimental community release intended for research, evaluation, and workflow exploration.


🚀 2. MTP Benchmark: Qwen3.6-27B vs Qwopus3.6-27B-v2-MTP

  • Speed: Qwopus3.6-27B-v2-MTP reaches 10.46 overall tokens/sec, compared with 6.29 tokens/sec for Qwen3.6-27B.
  • Latency: total evaluation time drops from 14,901.69s to 6,487.81s, saving 8,413.88s across the full run.
  • Output shape: MTP produces 67,862 completion tokens versus 93,802 from Qwen3.6-27B, giving a more compact overall response profile.

[!IMPORTANT] Benchmark source: /workspace/renji-training/Jackrong/qwopus3.6-27B-v2-MTP/benchmark_27b_pair_report.md on the GB10 server. Local workspace date: 2026-05-22.


⚙️ 3. Test Environment & Configuration

  • Compute platform: GB10 dedicated server platform.
  • Evaluation format: same local GGUF server stack for both models.
  • llama-server total context: 49152.
  • Temperature / Top-p: 1.0 / 0.95.
  • Max generated tokens: no explicit cap; generation is bounded by the request budget.
  • Request format: /v1/chat/completions with user content as text payload.

📊 4. Full 30-Question Comparison


🧭 5. Domain Reading


  • Agentic coding and code review assistance.
  • DevOps runbooks, configuration generation, and incident diagnosis.
  • Multi-step math and probability derivations.
  • Structured reasoning with explicit intermediate logic.
  • Fast constrained output generation where latency matters.

Want more deterministic results?