Everything free. Upgrade when you're ready.
Every tool, every mode, every persona, 5-layer memory — every feature on every plan. You only pay for more credits.
New to credits? Learn how they work →
- 300 credits / mo≈ try the platform — upgrade or BYOK when ready
- BYOK — any provider
No account required with your own keys
Download- 5,000 credits / mo≈ daily coding sessions with tool use
- All managed models
No keys needed — we handle it
- 10,000 credits / mo≈ multiple sessions a day, tool-heavy workflows
- Extended context + early access
For power users
- 20,000 credits / mo≈ team-scale or all-day-every-day usage
- SSO + priority support
Team-scale usage
Credit top-ups
one-time750 credits
Quick boost
$3
$4.00 / 1K credits
2,000 credits
Best value
$8
$4.00 / 1K credits
4,000 credits
Power user
$15
$3.75 / 1K credits
Credits never expire — carry over month to month.
Have a coupon code?
What runs on each plan
Every plan gets every model and every feature. We only charge for credits — nothing is locked behind a paywall.
Free
300 credits per month
- Maestro— Qwen 3.7 Plus conductor — daily work, predictable cost
- Supernova— DeepSeek V4 Pro coordinator + V4 Flash specialists with Qwen builders — heavy multi-step work
- Aurora— Mistral three-tier fleet — Large 3 coordinator, Medium 3.5 builders/specialists, Small 4 intent gate. EU-only, open weights, GDPR-strict, sovereign infrastructure
- Creative Studio— image (Wan), audio, speech, video (MiniMax)
Paid
More credits — same features
- Maestro— Qwen 3.7 Plus conductor — daily work, predictable cost
- Supernova— DeepSeek V4 Pro coordinator + V4 Flash specialists with Qwen builders — heavy multi-step work
- Aurora— Mistral three-tier fleet — Large 3 coordinator, Medium 3.5 builders/specialists, Small 4 intent gate. EU-only, open weights, GDPR-strict, sovereign infrastructure
- Creative Studio— image (Wan), audio, speech, video (MiniMax)
BYOK — any plan
Your keys, no allowance consumed
- Qwen — unlocks Maestro mode
- Qwen + DeepSeek — unlocks Supernova mode
- Mistral — unlocks Aurora mode
- Anthropic (Claude)
- Moonshot (Kimi)
- Zhipu (GLM)
- MiniMax — required for Creative Studio
- or any OpenAI-compatible endpoint
Cost Calculator
Pick a model, pick how you work, see what it actually costs.
Ava is agentic, not a chatbot. One task = many tool calls + specialist handoffs + accumulating context. This calculator models real task cost, not per-message cost.
Included on managed plans
BYOK — your keys, any plan
One task = one intent (e.g. “fix this bug”) — whatever tool calls happen underneath.
Typical task size
Mode mix
Managed
Maestro
Qwen ensemble · free-tier default
Estimated monthly
$6.05
Tasks
240/mo
Usage
14.4M tok
Per task
60k tok
Where your 14.4M tokens go
file reads, grep/glob, bash output
system prompts, specialist handoffs
reasoning + text you see
Input cost
$2.30
11.5M × $0.20/M
Output cost
$3.74
2.9M × $1.30/M
Your 14.4M projected usage fits on the Pro plan at $19/mo (15M token allowance) — no per-token charges on top.
Frequently Asked Questions
What models are included on managed plans?
Every plan — Free included — gets the three orchestrated modes: Maestro (Qwen 3.7 Plus conductor for daily work), Supernova (DeepSeek V4 Pro coordinator + V4 Flash specialists with Qwen builders for heavy multi-step work), and Aurora (Mistral-only EU stack for GDPR-strict and sovereign workloads). Plus Creative Studio for image, audio, speech, and video. Free and Paid share identical features — we only charge for more credits, nothing is locked. With your own API keys, you can add raw models from Claude, Kimi, Zhipu, or any OpenAI-compatible provider on top.
Why three modes instead of a model list?
You pick the work, the fleet falls out of that. Maestro = predictable daily cost on a single Qwen conductor. Supernova = a frontier-tier coordinator (DeepSeek V4 Pro) directing Qwen specialists for heavy multi-step work — frontier reasoning at specialist prices. Aurora = entire stack stays inside the EU on Mistral weights, GDPR-strict, sovereign by default. The picker asks "what kind of work, what residency, what cost shape" — not "which name from the list".
Can I use the modes with my own API keys?
Yes — BYOK gets the same three modes, gated by the keys each fleet needs. Add a Qwen key to unlock Maestro. Add Qwen + DeepSeek to unlock Supernova. Add Mistral to unlock Aurora. Plus you can pick raw individual models directly (Claude, Kimi, GLM, etc.) per the keys you've configured. BYOK requests don't consume your credit allowance.
Do I need an account to use Ava?
No. The Free tier works entirely with your own API keys and doesn't require an account. Just install the extension, add an API key from any provider, and start working. You only need an account for managed plans.
Can I bring my own API keys on a paid plan?
Yes. Even on Pro, Ultra, or Enterprise, you can use your own API keys alongside your credit allowance. BYOK requests don't consume credits. Best of both worlds.
What happens if I run out of credits?
Purchase top-ups anytime without changing your plan. Top-up credits don't expire — they carry over until used.
How do credits work?
One credit is a unit of value, not a raw token count. Simple actions like a chat turn cost 2 credits; light routing or classification costs 1. Heavier work costs more — a full orchestrated task is around 10 credits, image generation is 12, music 50, video 150. Creative Studio voice is 10. Teach mode bills as standard chat turns (2 credits). Premium reasoning models add a multiplier on top — V4 Pro costs 6× the base. Cache-hit responses charge less automatically. You see the credit counter tick down in real time.
How does the single credits pool work?
Simple — one pool, one number. Every managed request (Maestro / Supernova / Aurora chat, plus Creative Studio) deducts from your credit allowance. The allowance resets each billing period. No separate pools, no confusion.
Can I donate to support the project?
Yes! If you use your own API keys and love what we're building, support us through GitHub Sponsors. Every contribution helps keep Ava open-source and free for everyone.
Can I cancel anytime?
Yes. No contracts, no commitments. Cancel anytime from your dashboard.
Powered by Qwen
Qwen 3.7 Plus is the conductor — 1M context, native function calling, Terminal-Bench #1 for agentic coding. Auto Mode routes across the Qwen family to match the task. Creative Studio runs on MiniMax. Frontier performance at a fraction of the cost, with a 50% enterprise pricing partnership baked in.
Support Ava
Using your own API keys and love what we're building? Your support keeps Ava open-source and helps us add new providers and features for everyone.
Local-first is sacred
Every feature works without an account. Every conversation, every memory, every piece of generated content lives on your machine first. Cloud sync is additive — never the master.
BYOK users stay fully local. Zero platform storage, zero telemetry, zero reporting. If privacy is the point, BYOK is the path.