What can I do with VelATL today?

Chat with named models, question your PDFs and office files, research the live web with citations, draft writing and work through code - plus spend the same credits through an OpenAI- and Anthropic-compatible API and an MCP server.

New accounts start with 1,000 free credits. AI replies consume credits at per-model token rates; research and images have flat published fees. Credits stay in your balance until spent and never expire.

Start free with 1,000 credits. Paid plans start at Rs 199/month (annual billing saves two months), and one-time top-ups work alongside any plan.

Honest comparison

Groq direct or VelATL on top?

Our fastest models run on Groq's hardware, so going direct is the cheapest per token by definition. This page is about everything around the tokens: billing currency, workspace, caps, and one ledger.

Stay on Groq when…

You want the lowest per-token price on one model, Groq's free tier for prototyping, and console rate limits are enough. Nothing we sell beats their raw price - it can't, we pay it.

Choose VelATL when…

You want UPI/INR billing with GST invoices, a web workspace plus OpenCode, Claude Code and MCP on the same credits, per-key spend caps, automatic refunds, and published cost margins.

What changes technically

VelATL's Swift and Balanced slugs are gpt-oss-20b and gpt-oss-120b served via Groq - same hardware family, same speed class. The API shape stays OpenAI-compatible either way:

The switch is a baseURL + model name

# Groq direct
base_url="https://api.groq.com/openai/v1", model="open-weights/gpt-oss-20b"

# VelATL (OpenAI SDK, OpenCode, Cline, Continue - unchanged otherwise)
base_url="https://velatl.ai/api/v1", model="fast"  # or "balanced"
api_key="velatl_..."  # Settings -> API keys, new keys capped at 5,000 cr/mo

Streaming, tool calling and JSON mode work on both; VelATL serves text + tools only (limits in docs).
VelATL adds per-key monthly spend caps counting settled + in-flight spend, per-request ledger entries, and automatic refunds on provider failure.
Out of credits → 402 insufficient_credits with a billing link - no silent card charges, because spend is prepaid credits, never postpaid.

How to compare the price honestly

Don't compare headlines - compare your mix. Take last month's input/output tokens per model, price them at Groq's list plus your card's forex markup, and compare against VelATL's per-1k credit rates at 100 credits ≈ ₹1 plus the plan fee. Our cost leaderboard publishes what you pay versus what inference costs us, so you can check the spread instead of trusting it.

Indian developers should also read AI API for India (UPI, GST, DPDP) and the VelATL vs OpenRouter comparison before deciding.

Start free with 1,000 credits See live model rates