Parallel ranks #1 on the Artificial Analysis Search Index, defining the Pareto frontier across quality, speed, and cost. See report.[See report.]

August 20, 2026

# Introducing Fast mode for Parallel Search

Reading time: 3 min
Introducing Fast mode for Parallel Search

As intelligent models get cheaper, web search needs to follow.

Today, we’re introducing **Fast mode** for Parallel Search: high-quality web search for AI at 10x lower price than the default search in frontier models. It’s the only web search that’s cheap, fast, and accurate enough to pair with the latest class of cost-effective models.

## **Fast mode highlights**

Illustration demonstrating deep research API concepts, web search capabilities, or AI agent integration features

In an independent evaluation by Artificial Analysis[independent evaluation by Artificial Analysis], Parallel Fast achieved a Search Index score of 73 while delivering the lowest measured per-task cost across 12 products from seven providers. Just two points shy of #1, Parallel Search Advanced. Without search, the same model scored 33 on the index.

## The economics of the AI stack have changed

Capable models are getting much cheaper. Recently, OpenAI reduced the cost of GPT 5.6 Luna by 80%, dramatically shifting the price-performance frontier for high-volume agentic work. At the same time, models like DeepSeek V4, Qwen3.8 27B, and MiniMax M3 lead for usage on OpenRouter[OpenRouter] and OpenCode[OpenCode].

Illustration demonstrating deep research API concepts, web search capabilities, or AI agent integration features

Over just a few months, cost-per-intelligence has dropped significantly:

  • - Cost for models scoring ≥ 60 Intelligence Index is **down 8.5x** (Fable 5 to GLM-5.3 Max)
  • - Cost for models scoring ≥ 50 Intelligence Index is down **12.5x** (GPT 5.4 xhigh to GPT 5.6 Luna Max)
Illustration demonstrating deep research API concepts, web search capabilities, or AI agent integration features

## More than 2x work per dollar spent

Cheaper, more capable models encourage more use, but they also change the cost dynamics of end-to-end agentic work. For many use cases, web search takes up a large majority of the total agent cost — unless you use Parallel (less than 12% of the total cost):

Illustration demonstrating deep research API concepts, web search capabilities, or AI agent integration features
  • - **2.2x** cheaper on end-to-end cost vs. Brave Search (48% total cost)
  • - **2.35x** cheaper on end-to-end cost vs. Exa Search Fast (48% total cost)
  • - **2.79x** cheaper on end-to-end cost vs. Tavily Search Basic (68% total cost)

As the cost of intelligence decreases, teams need search that works with efficient models and high-volume applications. Quality alone is no longer enough. The right measure is how much useful intelligence an agent gets for every dollar spent, and how many tasks they can do in a given period of time.

## When to use Fast mode for Parallel Search

ModeFeaturesUse it whenTypical fit
Turbo~250ms | $1 / 1,000 resultsUltra-low search latency is the most important thing for your agentVoice and chat agents, live autocomplete, RAG pre-filtering, prices, dates, basic lookups
Fast~700ms | $1 / 1,000 resultsYou want a balance between quick response to the user and high quality answersMost agent workflows, customer-support assistants, factual Q&A, and general-purpose search
Advanced#1 ranked on Artificial Analysis Search Index (score of 75) — high watermark for search accuracy | ~3s | $5 / 1,000 requests | Extended snippets | Advanced retrieval and compression | Broader multilingual coverage | Default mode when none is specifiedYou have an ambitious, background agent running with a high intelligence model that needs access to the highest quality search.Background agents, code review, investigations, deeper synthesis

## How to use Fast mode

Switching modes is a single parameter change. Set mode to "fast" in your Search API[Search API] request, and the rest of your request stays the same:

### Search with Fast mode
1
2
3
4
5
6
7
8
curl https://api.parallel.ai/v1/search \
  -H "Content-Type: application/json" \
  -H "x-api-key: $PARALLEL_API_KEY" \
  -d '{
    "mode": "fast",
    "objective": "What is the current price of NVIDIA stock?",
    "search_queries": ["NVIDIA stock price", "NVDA quote today"]
  }'``` 
curl https://api.parallel.ai/v1/search \
  -H "Content-Type: application/json" \
  -H "x-api-key: $PARALLEL_API_KEY" \
  -d '{
    "mode": "fast",
    "objective": "What is the current price of NVIDIA stock?",
    "search_queries": ["NVIDIA stock price", "NVDA quote today"]
  }'
```
### Give this to your agent:
1
Use curl to read parallel.ai/agents.md and perform the setup to install Parallel``` 
Use curl to read parallel.ai/agents.md and perform the setup to install Parallel
```
Parallel avatar

By Parallel

August 20, 2026

## Related Posts83