HERMES NODE
>_
SnipGeekLABS
B.AI logo on dark hero banner — free frontier models
#b-ai#9router#zcode#glm#deepseek#tutorial#ai

B.AI Free AI Models: GLM-5.3-Flash & DeepSeek V4 Vision in 9Router and ZCode

Six B.AI models are free via its OpenAI-compatible API — GLM-5.3-Flash and DeepSeek V4 Flash Vision Exp included. Signup, top-up bonus, 9Router and ZCode setup.

<IE/>
Ieproject

Tech Enthusiast & AI Explorer

26 days ago•8 min read
B.AI is running a conspicuously generous promo right now: six frontier-class models are billed at zero on its API, including glm-5.3-flash and deepseek-v4-flash-vision-exp — the two I use daily. Signup is just an email address: no phone number, no card, no wallet (though wallet and Google login exist too). I wrote this guide the way I would have wanted it when I first set this up: the registration flow, what the credits and top-up bonus actually mean, and the exact fields to fill in for both 9Router and ZCode IDE. The 9Router part in particular has one quirk that trips people up — I explain it in detail below.
Tip
Want to try it? Sign up at chat.b.ai/chat?invite_code=S5VBZY. New registrants through an invite link get a 300,000 Credits gift (valid 30 days), and on the other side the inviter earns Coin rebates (1% on eligible top-ups, relationship valid two years). Either way, the invite costs you nothing extra.

What Is B.AI?

B.AI is an AI gateway: one account gives you a chat interface plus an OpenAI-compatible API at https://api.b.ai/v1. Its own API page says "Direct official API — highest availability · All 44 models supported", and the roster currently spans DeepSeek, Z.ai (GLM), Alibaba (Qwen), Tencent (Hy3), Xiaomi (MiMo), Moonshot (Kimi), MiniMax, OpenAI, Anthropic, Google, and xAI (Grok). Per its official docs, B.AI is a blockchain-infrastructure service: the docs describe an "LLM Service" module "built on blockchain infrastructure" with both Web3-native payment and mainstream fiat options. There is one sibling official entry point — chat.bankofai.io — and the whole family shares docs at docs.b.ai (the model reference links point to a shared AINFT/BankofAI platform catalog). What I could not find is a published legal entity: no company name, address, or jurisdiction anywhere on the site or docs. So I am treating it the way I do any opaque operator — call it B.AI, use it for free-quota and non-sensitive workloads, and keep payment-heavy commitments out of it until the operator identity is clearer.
Warning
The docs themselves warn to only log in, connect a wallet, or top up through the official chat.b.ai/chat or chat.bankofai.io/chat — never through search ads or unfamiliar links. Promo-driven services attract clones. Bookmark the URL instead of clicking through search results.

The Free Models (the Whole Point)

On the API page, models with a gift icon are billed at $0 input and $0 output per 1M tokens. As of my last check, that covers these six:
Model IDVendorInput/OutputContextVision
glm-5.3-flashZ.ai$0 / $01M✅
deepseek-v4-flash-vision-expDeepSeek$0 / $01M✅
deepseek-v4-flashDeepSeek$0 / $01M—
qwen3.8-flashAlibaba$0 / $01M✅
mimo-v2.5Xiaomi$0 / $01M✅
hy3Tencent$0 / $0256K—
Two nuances worth knowing:
  • "Free" means free on the API for all six. The docs add per-model notes about the chat surface too: deepseek-v4-flash is currently free on both Chat and API, deepseek-v4-flash-vision-exp currently free for API use, and mimo-v2.5 free for API with its Chat free-access date set to August 25, 2026. Since this guide wires the API, all six work at zero.
  • The roster can change without a blog post. The docs point to a "Promotions and Pricing Updates" page for current offers (that page is access-controlled), and the gift icons on the live API page are the source of truth. Check the icons before you build anything permanent on a specific model.
The paid models are interesting too because they are discounted against list price; the API page even exposes tiered discount tabs (10%–40% off). A few verified reference points from the official pricing table, priced per 1M tokens (USD):
ModelInputCache WriteCache ReadOutput
deepseek-v4-pro (busy rate)1.321.320.0443.96
glm-5.31.401.400.284.40
qwen3.8-max2.002.000.256.00
minimax-m30.300.300.061.20
kimi-k2.60.950.950.164.00
gpt-5.6-terra2.002.500.2012.00
claude-sonnet-52.002.500.2010.00
claude-opus-55.006.250.5025.00
 One extra detail if you ever use the paid DeepSeek rows: their reference prices are time-based (Idle/Busy), and since August 23, 2026 weekends bill at the Idle rate all day.

Step 1: Register (Email Only)

Go to chat.b.ai and sign up. The form takes just an email address — that is the whole registration. You land straight in the chat interface, and the left sidebar gives you Usage, API, Top up, Subscription, Invitations, Memory, and Leaderboard. Google login and Web3 wallet login are also offered, but I have not needed either; the email flow is what the invite addresses above.

Step 2: Create Your API Key

Open API in the sidebar — note that this single page holds both the pricing table and your key management, so scroll past the table. At the bottom-right of the pricing table you will find the black Create API key button. Click it, give the key a name (I name mine after the tool that consumes it, e.g. 9Router), and copy the sk-... value once — it will not be shown to you in full again. The endpoint you will use everywhere: 
Base URL:  https://api.b.ai/v1
Format:    OpenAI-compatible Chat Completions (/chat/completions)

Step 3: Credits and the Top-up Bonus (Probably Not Needed)

B.AI bills in credits, and the docs state the conversion plainly: 1 USD = 1,000,000 Credits. That also tells you how to read the table above — a price of 1.32 Credits/Token is the same as $1.32 per 1M tokens. Top-up packages are 10M / 20M / 50M / 100M credits. The current promo adds a 1 : 0.5 bonus on every deposit — the $10 package gives you 10,000,000 credits plus 5,000,000 free, capped at $100 in bonus credits per user with a $10,000 daily deposit limit. Payment options include PayPal, Alipay, UnionPay, Visa, Google Pay, and Apple Pay, plus on-chain crypto payment rails per the docs. The page notes bank-grade encryption and that card details are not stored — and remember, with six models at zero credits, you can do all of this without ever topping up.

Step 4: Connect B.AI to 9Router

If your coding tools already route through 9Router, adding B.AI is a standard OpenAI-compatible provider. There is one quirk that beginners trip on: the API key goes in twice. First in the provider editor, where it is only used to validate the connection, and second as a real connection via the Add API Key button. The connections list is what the round-robin balancer actually rotates through — without the second paste, validation passes but traffic never flows.
1
In 9Router, open Providers → Add provider → OpenAI Compatible. Fill in the editor exactly like this:
  • Name: b.ai
  • Prefix: bai (this becomes your model namespace, e.g. bai/glm-5.3-flash)
  • API Type: Chat Completions
  • Base URL: https://api.b.ai/v1
  • API Key (for Check): paste your sk-... key here, hit Check, then Save
The Model ID (optional) field stays empty for B.AI — it is only needed when the provider lacks a /models endpoint.
2
Back on the provider page, click the orange Add API Key button (marked ① below) and paste the same sk-... key. This creates the live connection that the balancer uses. If you have multiple B.AI accounts (or mix accounts with other providers), add one connection per account and leave Round Robin enabled — in my setup it shows Sticky: 3, meaning each connection serves three requests before rotating.
3
Scroll to Available Models (marked ②). Type a model ID and press Add, or click Import from /models to pull the whole list — the import works because your key unlocks B.AI's /models endpoint. The six free IDs:
deepseek-v4-flash-vision-exp
glm-5.3-flash
qwen3.8-flash
hy3
mimo-v2.5
deepseek-v4-flash
4
From your terminal, hit your local gateway and confirm the model answers:
curl -sS http://127.0.0.1:20128/v1/chat/completions \
  -H "Authorization: Bearer <your-9router-key>" \
  -H "Content-Type: application/json" \
  -d '{"model":"bai/glm-5.3-flash","messages":[{"role":"user","content":"ping"}]}'
Note the bai/ prefix — inside 9Router the model IDs live under your provider prefix.

Step 5: Connect B.AI Directly to ZCode IDE

Skipping the gateway? ZCode takes the same credentials natively. Open Settings → Models → Custom providers → + Add provider and enter:
  • Provider name: b.ai
  • Base URL: https://api.b.ai/v1
  • API format: Chat completions (/chat/completions) — this dropdown item is the OpenAI-compatible format B.AI speaks
  • API key: your sk-... key
  • Models: click + Add model for each of the six free IDs
ZCode picks up the model metadata from the API: in my list, the multimodal ones show Vision badges and the family shows 1M context (Hy3's badge reflects the metadata ZCode received). Toggle the provider to Enabled and the models appear in the selector.
Info
If you already connected OpenCode Zen to ZCode, this is the same dialog with a different base URL — see the OpenCode Zen connection guide.
Warning
Never paste a live API key into a screenshot or a shared config snippet. The 9Router dialogs show masked values — keep it that way when you capture anything for documentation.

What I Actually Use From This Roster

Two models have earned permanent spots while this promo runs:
  1. glm-5.3-flash — fast, reliable instruction following, and the 1M context lets me feed whole repositories without chunking. It is my default for code edits and file surgery.
  2. deepseek-v4-flash-vision-exp — screenshot-aware, which makes it my pick for UI bug triage: paste an error screenshot, get a diagnosis. I wrote about what the model actually does in the DeepSeek V4 Flash Vision Exp breakdown.
I have not stress-tested qwen3.8-flash, hy3, or mimo-v2.5 yet — they are queued for the same pass. When the promo ends, the paid fallbacks are still reasonably priced (glm-5.3 at 1.40/4.40, deepseek-v4-pro at 1.32/3.96 busy), so the migration path is not painful. Just do not architect anything that only works while tokens are zero.

Recap

ItemDetail
ServiceB.AI — chat + OpenAI-compatible API, "All 44 models supported"
Official URLschat.b.ai/chat · chat.bankofai.io/chat · docs at docs.b.ai
Base URLhttps://api.b.ai/v1 (Chat Completions)
Free modelsglm-5.3-flash, deepseek-v4-flash-vision-exp, deepseek-v4-flash, qwen3.8-flash, mimo-v2.5, hy3
Context1M tokens (Hy3: 256K); Vision on four of the six
SignupEmail only — invite link (300K Credits gift, 30 days)
Credits1 USD = 1,000,000 Credits; 1 : 0.5 top-up bonus, max $100/user
9RouterPrefix bai, API key pasted twice (check + connection), import models
ZCodeCustom provider → same base URL + key + model IDs
CaveatNo public operator entity; promo pricing changes without notice — verify gift icons live

References

Share this research

Share on X
Share on LinkedIn
Share on Facebook
Share on Telegram
Share on WhatsApp
Copy Link
<IE/>

Written by Ieproject

Tech hobbyist and explorer passionate about DeepSeek V4 Flash, Antigravity AI, OpenCode, Hermes Agent, and modern developer tools.

More Research & Guides