Skip to content

models.mdx

Latest commit

 

History

History
216 lines (151 loc) · 6.82 KB

models.mdx

File metadata and controls

216 lines (151 loc) · 6.82 KB
title AI Models & Pricing
description Available AI models on Agentbot with transparent pricing. BYOK — pay providers directly, zero markup.

<img src="https://indigo-decent-condor-546.mypinata.cloud/ipfs/bafybeigkpl3kax3x5wpx4xyyfldhyq6hqcwlihz5ku4cxc4ltufow4osyi" alt="Agentbot models" height="360" style={{borderRadius: '12px', width: '100%', objectFit: 'cover', marginBottom: '24px'}} />

Default model: MiMo-V2-Pro

MiMo-V2-Pro is Xiaomi's flagship AI model and the default model for all new Agentbot container configurations.

  • Top-ranked in programming benchmarks — Excellent for agent and coding tasks
  • 1M context length — Large context window for complex workflows
  • Cost-effective — Strong performance at competitive pricing
  • Stable — Proven reliability on OpenRouter infrastructure

MiMo-V2-Pro pricing

Tier Input Output
Standard $1/M $3/M

Specs

Spec Value
Total Context 1M tokens
Provider Xiaomi (via OpenRouter)
Model ID openrouter/xiaomi/mimo-v2-pro

Gemini Flash 1.5 was the previous default model. Existing agents provisioned with Gemini Flash 1.5 continue to use it unless you change the model in your agent settings. New agents are provisioned with MiMo-V2-Pro.

Supported Models

Agentbot supports 300+ models via OpenRouter. BYOK (Bring Your Own Key) — you pay providers directly with zero markup.

Recommended Models

Model Input Output Best For
MiMo-V2-Pro $1/M $3/M Default — agents, coding, long context
Claude Sonnet 4 $3/M $15/M Complex reasoning, analysis
GPT-4o $2.5/M $10/M General purpose, vision
Gemini 2.5 Flash $0.15/M $0.60/M Fast, cost-effective
Gemini Flash 1.5 $0.075/M $0.30/M Fast, cost-effective
DeepSeek R1 $0.55/M $2.19/M Reasoning, math
Solana Agent Kit Via OpenRouter Via OpenRouter DeFi, NFTs, token operations (60+ Solana actions via MCP)

Free Models

Model Cost Notes
Gemini 2.0 Flash Free 150 RPM limit
Gemma 3N Free Good for testing

Full Pricing Table

Model Input (per 1K tokens) Output (per 1K tokens)
Gemini 2.0 Flash Free Free
Groq Llama 3 £0.0002 £0.0002
Gemini 1.5 Flash £0.0001 £0.0005
MiMo-V2-Pro $1/M $3/M
GPT-4o Mini £0.0003 £0.0012
Claude 3 Haiku £0.0002 £0.0010
GPT-4o £0.0022 £0.0088
Claude Sonnet 4 $3/M $15/M
DeepSeek R1 $0.55/M $2.19/M

Token quotas

Each plan includes a monthly token allowance. Agentbot tracks your cumulative token usage for the current calendar month and rejects requests that would exceed your plan limit. The quota resets automatically at the start of each month.

Plan Monthly token limit
Solo 2,000,000
Collective 6,000,000
Label 20,000,000
Network Unlimited

When you exceed your quota, chat completion requests return a 429 status with the QUOTA_EXCEEDED error code:

{
  "error": "Monthly token quota exceeded for plan \"solo\". Used 2,000,000 of 2,000,000 tokens. Quota resets at the start of next month.",
  "code": "QUOTA_EXCEEDED"
}

To continue using AI features before the month resets, upgrade to a higher plan from the billing page.

You can check your current quota usage from the dashboard cost API. The response includes a quota object with usedTokens, percent, and an overageWarning flag that turns true at 80% usage.

If Agentbot cannot reach the usage database, quota enforcement fails open — your request proceeds without a usage check. This ensures temporary infrastructure issues do not block your agents.

Plan model access

The public plans unlock different model sets:

Plan Models Available
Solo (£29) MiMo-V2-Pro, Claude Sonnet 4
Collective (£69) + Gemini 2.5 Flash
Label (£149) + DeepSeek R1, Solana Agent Kit

Custom and white-label deployments can unlock broader model access by arrangement, but the public self-serve plans are Solo, Collective, and Label.

Solana Agent Kit

The Solana Agent Kit model provides 60+ on-chain Solana actions through MCP (Model Context Protocol). It is available on the Label plan.

Capabilities

  • DeFi — Token swaps, liquidity provisioning, yield farming
  • NFTs — Minting, listing, and managing NFT collections
  • Token operations — Token creation, transfers, and balance queries
  • On-chain data — Market data, transaction history, wallet lookups

Specs

Spec Value
Model ID openrouter/solana/solana-agent-kit
Provider Solana (via OpenRouter)
Required Plan Label
Actions 60+ Solana on-chain actions

Configuration

{
  "models": {
    "default": "openrouter/solana/solana-agent-kit"
  }
}

The Solana Agent Kit requires a Label plan. If you are on a Solo or Collective plan, upgrade to Label to access this model.

BYOK (Bring Your Own Key)

Agentbot is BYOK — you connect your own API keys from AI providers. We charge zero markup on model usage.

Supported Providers

  • OpenRouter — 300+ models, one key
  • Anthropic — Direct Claude access
  • OpenAI — Direct GPT access
  • Google — Direct Gemini access
  • Ollama — Local models, free

Getting an OpenRouter Key

  1. Go to openrouter.ai
  2. Create an account
  3. Go to Keys → Create Key
  4. Add credits (minimum $5)
  5. Copy key to Agentbot dashboard

Key Security

  • API keys are encrypted at rest
  • Never shared with third parties
  • Can be rotated at any time
  • Deleted on account removal

Model Selection

From Dashboard

  1. Go to Settings → Models
  2. Select your preferred model
  3. Your agent uses it for all new conversations

From Config

{
  "models": {
    "default": "openrouter/xiaomi/mimo-v2-pro",
    "fallbacks": ["openrouter/anthropic/claude-sonnet-4", "openrouter/google/gemini-2.5-flash"]
  }
}

Per-Conversation

Override the default model for specific conversations via the chat interface.

FAQ

Q: Why MiMo-V2-Pro as default? A: Top-ranked in programming benchmarks with a 1M context window — ideal for agent and coding tasks at a competitive price point via OpenRouter.

Q: Can I use a different default? A: Yes — change it in Settings → Models anytime.

Q: Do you mark up model prices? A: No. Zero markup. You pay providers directly at their rates.

Q: What if my model goes down? A: Fallback models automatically kick in. Configure fallbacks in your agent settings.