Free
$0
No card required
Try it on your own work.
Start free- Up to 5 agents
- Light usage, to try things out
- Bring your own model, free for good
- 1 GB workspace a month
- 1 GB hub storage a month
- 30 min web browsing a month
- Community support
One price per person or per organisation. Models at the provider’s price.
One person. Usage included.
Prices include VAT. EU businesses with a VAT ID and customers outside the EU pay the net price.
Per organisation. Unlimited users and agents.
Models: bring your own provider, or use ours at the provider’s price. More at the usual rates: web browsing $0.08 an hour, gpu $2.00 a GPU-hour, storage $0.0605 per GB a month.
By quote
A yearly licence for your cloud, your servers or an air-gapped network, priced after a call about your setup.
Contact Salesz-ai/glm-5.3, input price per million tokens. The router compares them on every request.
| Provider | Input | First token |
|---|---|---|
| llm.buzz | 0.98 | 6.17 s |
| Prime Intellect | 0.98 | 0.58 s |
| Alibaba | 1.19 | 2.07 s |
| Novita | 1.4 | 1.72 s |
| Scaleway EU | 2.0743 | 0.28 s |
First token is the 24-hour median. From llm.buzz, 2026-10-27.
Serve models? List them on llm.buzzYes. Upgrades take effect immediately and we prorate the difference. Downgrades take effect at the next billing cycle. No fees either way.
Every paid plan includes a monthly usage allowance: Standard is the baseline, Pro gives you 8x that and Max 20x, all on Surogate and Surogate Pro. There are no overage bills. When the allowance runs out, agents pause until you top up your wallet (from $2, and the cash rolls over) or the next cycle starts.
The operation stops. No auto-upgrade, no overdraft, no end-of-month invoice. Top up the wallet with any amount (top-up balances never expire) and the task resumes. You can also bring your own model provider and pay it directly, which bypasses the allowance entirely.
Each consumable resource (usage, web browsing time, compute credits) works like a prepaid wallet. Your plan grants a monthly balance that resets each cycle. Top-ups add more on demand, and that cash rolls over as long as your account is active. When the wallet hits zero, the operation stops.
Yes. Connect any OpenAI-compatible provider: OpenRouter, OpenAI, Anthropic, Together, Groq, vLLM, TGI, Ollama or your own fine-tuned deployment. You pay them directly. In settings, cancel the bundled usage portion of your plan to save the cost.
Standard, Pro and Max are single-user. For a team, use a Teams plan: one organisation with unlimited users and agents. Scale adds single sign-on, roles and audit logs.
Two model levels on every plan: Surogate (Sonnet class) and Surogate Pro (Opus class). Pro is stronger and draws about 2.5x faster from the allowance.
Your workspace is private and yours, and hub artifacts are private by default. We do not train on your data. You can export everything anytime, with no egress fees on hub content.
Tokens the model consumed are billed whether the agent succeeded or not, as with every model-based service. We do not charge browser time or extra compute for failures on our end.
The Free plan is open-ended, with a starter usage allowance or your own model from day one. If a paid plan does not work for you in the first 14 days, we refund it on request.
Yes. Agents run continuously within your plan’s limit on agents working at once. Usage and web browsing hours are the wallets to watch on long tasks, and the dashboard shows live consumption.
Train on your own data on our managed GPU cloud: pick a GPU, start, and the run draws from your compute credits. Or connect your own cloud (AWS, GCP, Lambda, RunPod or your own Modal account) through dstack or skypilot and train there: your GPUs, your bill, no credits used and no GPU markup from us. Either way we orchestrate the run, capture logs and metrics, and track lineage in the hub.
Compute credits meter GPU time for fine-tuning, synthetic datasets and eval runs (HumanEval, MBPP, terminal-bench, SWE-bench, your own suites). The usage allowance covers your agents’ model calls. One credit is one minute on the baseline GPU tier, so faster GPUs draw more (an H100 minute is about 6.7 credits). Standard includes 180 credits a month (3 GPU-hours), Pro 900 (15), Max 2,400 (40). Top-ups are $2.00 per GPU-hour and roll over.
You may already have credits, a preferred region, compliance needs or a GPU provider. Our GPU cloud is the default so you can start without an account, but your provider can bill you directly instead, with no GPU markup from us and no credits used.
Yes. Deploy on your own cloud (dstack or skypilot) and we orchestrate the endpoint. Your agents call your models directly, with no usage charges from us.
A git-backed registry for your models, datasets and checkpoints, like a private Hugging Face Hub, backed by Cloudflare R2 with no egress fees. Private storage is 1 GB on Free, 10 GB on Standard, 50 GB on Pro and 200 GB on Max. Extra hub storage is $0.0605 per GB a month on any plan.
Add storage ($0.0605 per GB a month for workspace or hub) and top up the wallet, or move to a Teams plan for unlimited users. Write to us if you are not sure.
Partners pay no plan fee for their own workspace, and earn 30% of every client plan, every month, for as long as the client stays.