Startup Ideas Bank

A giant model is not a business plan.

AI roast score: 41/100 (F)

The idea

You did not pitch a company. You pitched a research demo with a price tag, a moat wish, and a global sales fantasy. The hard part is not making a big model. It is getting anyone to pay for it.

In the founder's own words

MoonshotAI/Kimi-K3 — Open Frontier Intelligence

📰   Tech Blog |     📄   Full Report

1. Model Introduction

Kimi K3 is an open-weight, native multimodal agentic model and our most capable model to date. It is a 2.8T-parameter model built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), with native vision capabilities and a 1-million-token context window. It is the world's first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning.

Key Features


New Architecture : Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), and scales up MoE sparsity with a Stable LatentMoE framework that activates 16 out of 896 experts — yielding an approximate 2.5× improvement in overall scaling efficiency over Kimi K2.

Long-Horizon Coding : Operating with minimal human oversight, Kimi K3 sustains long engineering sessions, navigates massive repositories, and orchestrates terminal tools — from GPU kernel optimization and compiler development to vision-in-the-loop game dev, CAD, and even chip design.

Agentic Knowledge Work : Kimi K3 advances end-to-end knowledge work, producing deep research with interactive visualizations, widgets and dashboards, and motion design and video editing, powered by its native multimodal architecture.

Native Multimodality & Long Context : Kimi K3 understands text, images, and video within the same model, and supports a 1-million-token context window.

Open Frontier Weights : We release the full Kimi K3 model weights under the Kimi K3 License, making frontier intelligence openly available for research, deployment, and further innovation.

2. Model Summary




Architecture
Mixture-of-Experts (MoE)


Total Parameters
2.8T


Activated Parameters
104B


Number of Layers
93


Number of Dense Layers
1


Attention-Layer Composition
69 KDA + 24 Gated MLA


Attention Hidden Dimension
7168


Number of Attention Heads
96


Latent MoE Dimension
3584


MoE Hidden Dimension (per Expert)
3072


Number of Experts
896


Selected Experts per Token
16


Number of Shared Experts
2


Vocabulary Size
160K


Context Length
1048576


Attention Mechanism
KDA & Gated MLA


Activation Function
SiTU-GLU


Vision Encoder
MoonViT-V2


Parameters of Vision Encoder
401M


Quantization
MXFP4 weights / MXFP8 activations (quantization-aware training)


Modality
Text, Image



3. Evaluation Results




Benchmark
Kimi K3 (max)
Claude Fable 5 (max, w/ fallback)
GPT-5.6 Sol (max)
Claude Opus 4.8 (max)
GPT-5.5 (xhigh)
GLM-5.2 (max)



Reasoning & Knowledge


GPQA Diamond
93.5
92.6
94.1
91.0
93.5
91.2


CritPt
23.4
28.6

The roast

This reads like a frontier model announcement, not a startup. You say “open-weight,” “1-million-token context window,” and “frontier intelligence,” but the only concrete buyer story is a mix of “enterprise, developers, professionals” and “premium” pricing. That is not a wedge. That is three buyer types and a bill. Your own profile makes the problem worse. You are a solo founder, you have no funding, you are at the idea stage, and your biggest unknown is capital. Meanwhile the thing you are proposing needs absurd compute, elite research talent, and a long sales cycle. That is not a mismatch. That is the business model screaming at you. The pitch also blends open weights, subscription, license, and usage revenue into one blob. That usually means nobody knows who pays, why they pay, or what stops them from switching to a stronger model next month. Open models win when distribution is obvious or the cost curve is brutal. You have neither proven here.

Red flags

Verdict

Pick one buyer, one job, and one economic reason to exist, or this stays a cool announcement and never becomes a company.

Start with your own startup idea →

More ideas roasted by AI

Browse all startup ideas →