Your team is probably overpaying for AI by 60%.
This dashboard shows you exactly how much — and what to do about it.
Moonshot AI's Kimi K3 (July 2026) matches GPT-4o on mathematical reasoning — at $0.40/1M input tokens vs $1.00/1M for GPT-4o.
Most engineering teams are defaulting to OpenAI out of habit, not analysis.
This tool fixes that.
No install. No signup. Open and use instantly.
Drag sliders for your monthly token volume → see exact monthly cost for every model in real time.
Example: 50M input + 20M output tokens/month
- Kimi K3: $52/month
- GPT-4o: $130/month ← 2.5× more expensive
- DeepSeek V3: $13/month ← cheapest option
Interactive Chart.js radar across: MMLU · HumanEval · MATH · GSM8K · ARC · HellaSwag · TruthfulQA · BBH
Side-by-side pricing, context window, strengths, and live benchmark bars for all 6 models.
6 pre-built routing strategies for real use cases — with % allocation per model:
- 🧠 Complex Reasoning & Math
- 💻 Code Generation & Review
- 📄 RAG / Document Analysis
- ✍️ Creative & Marketing Content
- 🔒 EU Regulated Enterprise
- 💰 Cost-Optimized at Scale
Scores each vendor on Regulatory Risk, Data Privacy, Export Control, and Vendor Lock-in — the factor Forbes says businesses are ignoring.
| Model | Org | Origin | Input $/1M | Output $/1M | Context | Open Source |
|---|---|---|---|---|---|---|
| ⚡ Kimi K3 | Moonshot AI | 🇨🇳 | $0.40 | $1.60 | 128K | ✅ |
| GPT-4o | OpenAI | 🇺🇸 | $1.00 | $4.00 | 128K | ❌ |
| Claude 3.5 Sonnet | Anthropic | 🇺🇸 | $0.80 | $2.40 | 200K | ❌ |
| Gemini 1.5 Pro | Google DeepMind | 🇺🇸 | $0.70 | $2.10 | 1M | ❌ |
| DeepSeek V3 | DeepSeek | 🇨🇳 | $0.14 | $0.28 | 64K | ✅ |
| Mistral Large 2 | Mistral AI | 🇪🇺 | $0.60 | $1.80 | 128K | ❌ |
| Benchmark | 🥇 Winner | Kimi K3 | GPT-4o | Claude 3.5 |
|---|---|---|---|---|
| MATH | Kimi K3 | 90.1 | 76.6 | 71.1 |
| MMLU | GPT-4o (88.7) | 88.5 | 88.7 | 88.3 |
| HumanEval | Claude 3.5 (93.7) | 87.2 | 90.2 | 93.7 |
| GSM8K | GPT-4o (97.5) | 95.3 | 97.5 | 96.4 |
| TruthfulQA | Claude 3.5 (80.3) | 72.5 | 77.8 | 80.3 |
Key insight: Kimi K3 is the #1 model for math and reasoning — the most common use case for LLM APIs — at a fraction of GPT-4o's price.
"The AI moat is no longer the model. It's the application layer — and the cost at which you can run it." — Forbes, July 2026
This dashboard is a direct, practical response to that article. It addresses all 5 business implications:
| Forbes Implication | This Dashboard |
|---|---|
| 1. Geopolitical risk is real | → Geopolitical Risk Matrix |
| 2. Model performance parity is here | → 8-benchmark radar |
| 3. Multi-vendor is the smart play | → Strategy Advisor with % routing |
| 4. AI tokenomics matter | → Real-time cost calculator |
| 5. ROI > benchmark obsession | → Hourly rate × savings = actual value |
📖 Full guide → INSTALL.md — includes all methods, feature walkthrough, how to add models, and deployment options.
TL;DR:
git clone https://github.com/karidasd/kimi-k3-intelligence.git
cd kimi-k3-intelligence
python -m http.server 3000
# open http://localhost:3000No npm install. No pip install. No environment variables. Pure static site.
| Technology | Why | |
|---|---|---|
| 📄 Core | Vanilla HTML/CSS/JS | Zero dependencies, instant load |
| 📊 Charts | Chart.js | Radar chart with real benchmark data |
| 🔤 Fonts | Outfit + JetBrains Mono | Premium, readable typography |
| 🎨 Icons | FontAwesome 6 | UI micro-icons |
| 🌐 Hosting | GitHub Pages | Free, instant, no config |
📖 Full guide → CONTRIBUTING.md
Found wrong benchmark data or a price change? PRs welcome. One rule: always include a source link (official pricing page, arXiv paper, or leaderboard).
# Quick contribution workflow:
git checkout -b update/model-name-pricing
# edit app.js MODELS array
git commit -m "update: GPT-4o input price $1.00 → $0.80 (source: openai.com/pricing)"
git push && open a PR
Built by DARKAIS Data Science · 2026
"Because some of the brightest ideas are born in the dark."