Multi-Model LLM Compare

Send one prompt to many language models at once and get every answer side by side in a single table. Compare GPT, Claude, Gemini, Llama, DeepSeek, Kimi, Qwen, GLM, and 300+ more models on quality, speed, and cost, without juggling separate accounts or API keys.

What you get

One row per model, with:

• The full text response from each model
• Prompt tokens, completion tokens, and total tokens
• Latency in milliseconds and the provider that served it
• Finish reason, status, and any error message

Use cases

• Pick the best and cheapest model for a task before you commit to it in production
• Benchmark answer quality, speed, and price across providers on your own prompts
• A/B test a system prompt across models
• Build evaluation datasets for prompt engineering

How to use

1. Enter your Prompt
2. List the Models to compare, for example openai/gpt-4o-mini and anthropic/claude-haiku-4.5. Leave it empty to use a default cross-provider set
3. Optionally add a System prompt, a Temperature, and a Max output tokens cap
4. Run the actor. Each model becomes one row in the Dataset tab, ready to sort by cost, speed, or token count

Input example

{
  "prompt": "Explain quantum entanglement to a 10 year old in 3 sentences.",
  "models": [
    "openai/gpt-4o-mini",
    "anthropic/claude-haiku-4.5",
    "google/gemini-2.5-flash",
    "meta-llama/llama-3.3-70b-instruct"
  ],
  "maxTokens": 300
}

Output example

{
  "prompt": "Explain quantum entanglement to a 10 year old in 3 sentences.",
  "model": "google/gemini-2.5-flash",
  "provider": "Google",
  "response": "Imagine you have two special coins that always land on the same side...",
  "promptTokens": 15,
  "completionTokens": 65,
  "totalTokens": 80,
  "latencyMs": 1001,
  "finishReason": "stop",
  "status": "success",
  "error": null
}

Pricing

This actor is usage-based — you pay in proportion to how much work each model call actually does, not a flat fee per response.

• Short answers from small models cost very little; long answers from premium models cost more. A quick comparison across a few small models is typically a fraction of a cent each.
• You are only charged for models that successfully respond. A model that errors out costs you nothing.
• The token counts in each row (prompt, completion, total) show how much each model produced, so you can see what drives your spend and pick the most efficient model for your task.
• Tip: set Max output tokens to cap how much each model can generate and keep your spend predictable — billing scales with the amount each model produces.

What it costs in practice

Price depends on which models you pick and how long their answers are. As a rough guide, here's the approximate cost per model response:

Answer length	Small & fast models (Gemini Flash, GPT-4o mini, Llama, DeepSeek)	Mid models (Claude Haiku, Kimi, GLM)	Premium models (Claude Sonnet/Opus, GPT-5, Gemini Pro)
Short (~150 words)	under $0.005	~$0.005–0.01	~$0.02–0.05
Long (~600 words)	~$0.01–0.02	~$0.02–0.04	~$0.06–0.15

A few example runs:

• Quick check — the 4 default models (Gemini Flash, GPT-4o mini, Claude Haiku, Llama) on a short answer → about $0.01–0.02 for the whole run
• One premium model, detailed answer — Claude Opus on a ~600-word response → about $0.12–0.15
• Broad comparison — 6 mixed models on medium answers → roughly $0.10–0.20 total

These are approximate — your actual cost tracks the token counts shown in each row. Lower Max output tokens to bring costs down, and remember you're only billed for models that actually respond.

Free tier

You can try this actor for free, with limits:

• Up to 3 models per run
• Up to 512 output tokens per model
• 3 runs per day, with a 30-minute cooldown between runs

Upgrade to a paid plan to compare up to 12 models at once, set output length up to 8192 tokens, and run without daily limits or cooldowns.

Notes

• Each model runs independently, so one failing model never blocks the others. Failed models return a row with status set to error and a short message.