# Ampersand v0.1

> Ampersand v0.1 is humans&'s collaborator model, trained from GLM 5.3 for people to achieve goals together: it tracks what each person wants, flags conflicts, shares context, and asks questions in group chats. Through Vercel AI Gateway it takes text in and text out with reasoning effort and client-side function tools over a 1M-token context window.

**ID**: `humansand/ampersand-v0.1`  
**Creator**: humansand  
**Category**: text  
**Context**: 1M tokens  
**Pricing**: $1.00 / $5.20 per 1M · billed by your provider (BYOK)  
**Released**: 2026-10-20  
**Web page**: https://anyrouter.dev/model/humansand/ampersand-v0.1

**Input modalities**: text  
**Output modalities**: text  

**Capabilities**: chat, streaming, reasoning, function-calling

**Supported parameters**: max_tokens, temperature, stop, tools, tool_choice, reasoning, include_reasoning

## Usage

**Endpoint**: `POST https://anyrouter.dev/api/v1/chat/completions`

```bash
curl https://anyrouter.dev/api/v1/chat/completions \
  -H "Authorization: Bearer $ANYROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "humansand/ampersand-v0.1",
  "messages": [
    {
      "role": "user",
      "content": "Say hi in 3 words."
    }
  ]
}'
```

## Providers

| Provider | Input | Output |
| --- | --- | --- |
| vercel-byok | $1.00/M (list) | $5.20/M (list) |

BYOK routes are billed by your own provider key — AnyRouter charges $0. Rates marked (list) are the provider's published list price.

## Model card

- Architecture: mixture of experts [1]
- Total parameters: 744B [1]
- Active parameters: 40B [1]
- Native context: 1048576 tokens [1]
- License: Ampersand Terms of Service [1]

### Strengths [2,1]

- Leads the vendor's long-horizon Collaboration Over Time result (12.94), ahead of its GLM 5.3 base (9.89).
- Strongest teamwork result shown: 59.1% on Collaborative Intelligence at high effort.
- Top multi-agent coding results the vendor reports: 92.2% on SWE-bench Pro v2 HARD-51 and 77.9% on DeepSWE v1.1 in collaborative mode.

### Known limits [1,2]

- May misunderstand goals, give incorrect information, interrupt, or disclose inappropriate context; simulation results may not generalize to real groups.
- Higher interaction cost than leading models in the vendor's simulation: sometimes offers unwanted coaching when people just want an answer.
- Prompt-injection susceptibility stays close to its GLM 5.3 base in the vendor's tests.

### Use cases [1,2]

- Group-chat facilitation for workplace and social teams: surfacing context, flagging conflicts, and asking questions while tracking each person's goals.
- Multi-agent coding teams, with Ampersand as the lead agent coordinating subagents over long multi-step tasks.
- Long-running collaboration sessions that keep conversation state per group over months of simulated time.

### API notes [4,2]

- Owner's Agents API at https://ampersand.humansand.ai/v1: sessions via POST /agents/sessions, follow-up events streamed over SSE; one session per conversation keeps history.
- Reasoning effort low, medium, high, or max (default high today); up to 32,768 output tokens including reasoning.
- Function tools take JSON Schemas and are run by the caller's app; image input is rejected.

### Links

- announcement: https://ampersand.humansand.ai/blog/
- model card: https://ampersand.humansand.ai/blog/model-card.html
- docs: https://ampersand.humansand.ai/docs/index.html.md
- pricing: https://ampersand.humansand.ai/docs/PRICING.html.md
- Previous version: https://anyrouter.dev/model/z-ai/glm-5.3

### Sources

1. [Ampersand v0.1 Model Card](https://ampersand.humansand.ai/blog/model-card.html) (accessed 2026-11-08)
2. [Ampersand — Our first model trained for people to achieve more together](https://ampersand.humansand.ai/blog/) (accessed 2026-11-08)
3. [Ampersand Pricing and Limits](https://ampersand.humansand.ai/docs/PRICING.html.md) (accessed 2026-11-08)
4. [Build with Ampersand](https://ampersand.humansand.ai/docs/index.html.md) (accessed 2026-11-08)

## Source

- Upstream docs: https://ampersand.humansand.ai/blog/model-card.html