# win.sh

win.sh is an autonomous company operating layer for founders and operators. It connects to business tools, monitors company data, proposes the next move, and executes approved work inside user-defined rules, budgets, memory, and authority limits.

win.sh is not a generic chatbot. The product is built around one company operator per business, persistent company memory, visible approvals, cost tracking, and operating loops that can run on a schedule.

## AI Model Router

Canonical page: https://win.sh/ai-model-router
Router skill: https://win.sh/ai-model-router/SKILL.md
OpenAPI spec: https://win.sh/openapi.json
API OpenAPI spec: https://win.sh/api/openapi.json
Full model index: https://win.sh/router/models
Last model index update: 2026-06-29T20:14:04.498Z

Use the AI Model Router when an AI agent needs to choose a model by task, price, benchmark quality, speed, or fallback. Public reads require no auth.

### Quick Model Answers

- Cheapest useful model: Ministral 3 3B (ministral-3b-2512) at $0.100/1M blended tokens
- Fastest model: Ministral 3 3B (ministral-3b-2512) at 320 tokens/sec
- Smartest model: OpenAI: GPT-5.5 Pro (gpt-5.5-pro) with intelligence 100
- Best value model: DeepSeek: DeepSeek V4 Flash (deepseek-v4-flash) with value 1675.2

### Model Router Endpoints

- GET https://win.sh/router?task=code&latency=fast&budget=low&quality=high — recommended model, fallback, policy, and reason
- GET https://win.sh/router/raw?task=code&latency=fast&budget=low&quality=high — recommended provider-native model id as text
- Add providers=anthropic,openai to /router or /router/raw to limit recommendations to selected providers
- Add regions=us,eu,china to /router or /router/raw to limit recommendations by model provider region. Multiple regions are allowed.
- GET https://win.sh/router/cheapest — cheapest useful model as JSON
- GET https://win.sh/router/cheapest/raw — cheapest useful provider-native model id as text
- GET https://win.sh/router/fastest — fastest model as JSON
- GET https://win.sh/router/fastest/raw — fastest provider-native model id as text
- GET https://win.sh/router/smartest — highest benchmark score model as JSON
- GET https://win.sh/router/smartest/raw — highest benchmark score provider-native model id as text
- GET https://win.sh/router/best-value — best intelligence per dollar as JSON
- GET https://win.sh/router/best-value/raw — best value provider-native model id as text
- GET https://win.sh/router/models — full model table with methodology and sources
- Legacy /api/* aliases remain available for compatibility.

### Route By Task

- Code generation: GET https://win.sh/router?task=code&latency=fast&budget=low&quality=high
- Hard reasoning: GET https://win.sh/router?task=reasoning&latency=normal&budget=balanced&quality=high
- Cheap extraction: GET https://win.sh/router?task=extraction&latency=fast&budget=low&quality=balanced
- Summaries: GET https://win.sh/router?task=summary&latency=fast&budget=low&quality=balanced
- Planning: GET https://win.sh/router?task=planning&latency=normal&budget=balanced&quality=high
- Only Anthropic and OpenAI: GET https://win.sh/router?task=code&latency=fast&budget=low&quality=high&providers=anthropic,openai
- Only US or EU models: GET https://win.sh/router?task=code&latency=fast&budget=low&quality=high&regions=us,eu

### Curl Examples

```bash
curl "https://win.sh/router?task=code&latency=fast&budget=low&quality=high"
curl "https://win.sh/router?task=code&latency=fast&budget=low&quality=high&providers=anthropic,openai"
curl "https://win.sh/router?task=code&latency=fast&budget=low&quality=high&regions=us,eu"
curl "https://win.sh/router/raw?task=summary&latency=fast&budget=low&quality=balanced"
curl "https://win.sh/router/cheapest"
curl "https://win.sh/router/models"
```

### Current Top Models

1. DeepSeek: DeepSeek V4 Flash (deepseek-v4-flash, OpenRouter deepseek/deepseek-v4-flash) — value 1675.2, intelligence 72, $0.117/1M blended tokens, 210 tokens/sec, tags: fast, cheap, value
2. DeepSeek: DeepSeek V4 Pro (deepseek-v4-pro, OpenRouter deepseek/deepseek-v4-pro) — value 1590.1, intelligence 88, $0.566/1M blended tokens, 78 tokens/sec, tags: reasoning, value, math
3. Google: Gemini 3.1 Flash Lite (gemini-3.1-flash-lite, OpenRouter google/gemini-3.1-flash-lite) — value 518.4, intelligence 76, $0.625/1M blended tokens, 168 tokens/sec, tags: fast, cheap, long-context
4. Z.ai: GLM 5.2 (glm-5.2, OpenRouter z-ai/glm-5.2) — value 503.2, intelligence 86, $1.56/1M blended tokens, 92 tokens/sec, tags: open-weight, value, agent
5. OpenAI: GPT-5.4 Nano (gpt-5.4-nano, OpenRouter openai/gpt-5.4-nano) — value 497.1, intelligence 74, $0.515/1M blended tokens, 154 tokens/sec, tags: fast, cheap
6. Qwen: Qwen3.6 Flash (qwen3.6-flash, OpenRouter qwen/qwen3.6-flash) — value 479.7, intelligence 73, $0.469/1M blended tokens, 190 tokens/sec, tags: fast, cheap, long-context, open-weight
7. MoonshotAI: Kimi K2.7 Code (kimi-k2.7-code, OpenRouter moonshotai/kimi-k2.7-code) — value 464.9, intelligence 85, $1.57/1M blended tokens, 74 tokens/sec, tags: coding, agent, value
8. Qwen: Qwen3.7 Max (qwen3.7-max, OpenRouter qwen/qwen3.7-max) — value 420.5, intelligence 87, $2.00/1M blended tokens, 86 tokens/sec, tags: value, general, math, long-context, open-weight
9. Mistral Medium 3.5 (mistral-medium-3-5, OpenRouter mistralai/mistral-medium-3-5) — value 303.2, intelligence 82, $1.90/1M blended tokens, 104 tokens/sec, tags: fast, general
10. Google: Gemini 3.1 Pro Preview (gemini-3.1-pro-preview, OpenRouter google/gemini-3.1-pro-preview) — value 231.2, intelligence 92, $5.00/1M blended tokens, 64 tokens/sec, tags: long-context, reasoning, science, fast, cheap
11. Anthropic: Claude Sonnet 4.6 (claude-sonnet-4-6, OpenRouter anthropic/claude-sonnet-4.6) — value 196.4, intelligence 94, $6.60/1M blended tokens, 52 tokens/sec, tags: coding, agent, general
12. Anthropic: Claude Opus 4.8 (claude-opus-4-8, OpenRouter anthropic/claude-opus-4.8) — value 152.8, intelligence 99, $11.00/1M blended tokens, 28 tokens/sec, tags: frontier, reasoning, writing, science, math

### Router Method

- Scores use benchmark-style intelligence, coding, and reasoning fields.
- Model ids use provider-native ids when known. JSON responses include openRouterId separately when available.
- OpenRouter is used for live availability, token pricing, and context windows. It is not treated as the canonical provider identity source.
- Prices are normalized to blended dollars per 1M tokens.
- Value ranks useful intelligence per blended dollar after a quality floor.
- Task routing changes the weighting: code favors coding score, hard analysis favors reasoning, summaries and extraction favor cheap reliable execution, and fast tasks favor speed with quality guardrails.
- Provider allowlists are supported with providers=anthropic,openai. If no verified model matches the requested providers, the API returns 400 instead of routing outside the allowlist.
- Region allowlists are supported with regions=us,eu,china. If no verified model matches the selected regions, the API returns 400 instead of routing outside the allowlist.
- Every route returns a fallback when JSON is requested.

## Key URLs

- Homepage: https://win.sh/
- Pricing: https://win.sh/pricing
- How it works: https://win.sh/how-it-works
- AI Model Router: https://win.sh/ai-model-router
- AI Model Router skill: https://win.sh/ai-model-router/SKILL.md
- OpenAPI: https://win.sh/openapi.json
- Live activity: https://win.sh/live
- Integrations: https://win.sh/integrations
- Stripe integration: https://win.sh/integrations/stripe
- Plausible integration: https://win.sh/integrations/plausible
- GitHub integration: https://win.sh/integrations/github
- Slack integration: https://win.sh/integrations/slack
- Blog: https://win.sh/blog
- Docs: https://win.sh/docs
- API docs: https://win.sh/docs/api
- Support: https://win.sh/support

## Comparison Pages

- win.sh vs ChatGPT: https://win.sh/compare/chatgpt
- win.sh vs AI agents plus hiring: https://win.sh/compare/ai-agents-vs-hiring
- win.sh vs AutoGPT: https://win.sh/compare/autogpt
- win.sh vs CrewAI: https://win.sh/compare/crewai
- win.sh vs Make: https://win.sh/compare/make
- win.sh vs Zapier: https://win.sh/compare/zapier

## Audience

win.sh is for solo founders, SaaS operators, agencies, ecommerce operators, and small teams that want recurring business work to keep moving without hiring a larger operations team.

## Facts

- Product name: win.sh
- Brand casing: always lowercase
- Category: autonomous company operating layer
- Primary promise: run an autonomous company
- Core controls: approvals, budgets, memory, authority rules, traces
- Model router keyword: ai model router
- Model router page: https://win.sh/ai-model-router
- Machine-readable API spec: https://win.sh/openapi.json
- Update date: 2026-06-28