K2 Horizon — Model Drop — Dropday.ai

Model Drops Leaderboard

Sign in

← All model drops

Live drop

K2 Horizon

Institute of Foundation Models (MBZUAI)Dropped Sep 3, 20265 builds5 makers1 prompt

The latest builds made with K2 Horizon.

Browse in feed Submit a build

The most interesting things people are building with AI, all in one place — followed and tracked as they evolve from first demo to real product.

1379 indexedPress / to search--:--:-- UTC

Dropday is new — we're looking for genuine feedback.

Home Leaderboard Top creators What is this?Privacy Terms

DROPDAY.ai

Qwopus3.8-27B-Flash

Qwen3.8-27B-Escha-W2 GGUF

vLLM for AMD (Windows)

K2 Horizon family

3 ways to build

One generation, tuned for different levels of capability, speed, and cost.

Read the official Institute of Foundation Models (MBZUAI) release

K2 0.9B
K2 Horizon 0.9B is the smallest model in IFM's fully open fleet of six LLMs (weights, code, training data, and methodologies released under Apache 2.0).
0 builds
K2 3.7B
K2 Horizon 3.7B is part of IFM's fully open six-model fleet released September 3, 2026.
0 builds
K2 7B
K2 Horizon 7B belongs to IFM's fully open fleet of six models (Apache 2.0, with weights/code/data).
5 builds

Worth opening first

Standouts

Qwopus3.8-27B-Flash

Qwopus3.8-27B-Flash

A fine-tuned Qwen3.8-27B optimized for fast, efficient agentic reasoning and tool-use.

Qwen3.8-27B-Escha-W2 GGUF

PlayableOpen source

Qwen3.8-27B-Escha-W2 GGUF

A high-performance llama.cpp port of Escha's 2-bit quantization using custom CUDA kernels.

vLLM for AMD (Windows)

vLLM for AMD (Windows)

High-performance vLLM optimization for AMD hardware on Windows

Same brief, other models

All challenges →

Challenge

Flappy Bird

Same brief across models — older takes on the left, newer on the right.

40 / 49Compare all →

Playable3 Prompts1 More builds1

Playable

3

Qwopus3.8-27B-Flash

Qwopus3.8-27B-Flash

A fine-tuned Qwen3.8-27B optimized for fast, efficient agentic reasoning and tool-use.

Apps

Qwen3.8-27B-Escha-W2 GGUF

PlayableOpen source

Qwen3.8-27B-Escha-W2 GGUF

A high-performance llama.cpp port of Escha's 2-bit quantization using custom CUDA kernels.

Apps

Experiential Labs AI Platform

Experiential Labs AI Platform

Build simulations from traces, evaluate models, and serve optimized endpoints.

Apps

Prompts

1

Flappy Bird

A mobile-responsive Flappy Bird clone built in a single prompt.

Games

More builds

1

vLLM for AMD (Windows)

vLLM for AMD (Windows)

High-performance vLLM optimization for AMD hardware on Windows

Apps

Built something with K2 Horizon?

Paste the URL and Dropday.ai will credit you, track the build as it evolves, and put it on this drop.