Bee
A curated model catalog and AI gateway from qialike — endpoints, authentication, pricing, privacy and teams.
Bee is qialike’s AI gateway: a curated list of tested and verified models served behind one
API key and one base URL, https://qialike.com/bee. It speaks the OpenAI, Anthropic and Google
protocols, so any compatible tool can call it — and it is completely optional. You do not need
Bee to use qialike, and qialike does not have to be the client.
Only a few models work well as coding agents, and providers serve them very differently. We tested a select group, worked with the providers on how they were served, and benchmarked the combinations before recommending them.
Background
There are a great many models on the market, but only a few of them are actually suitable for use as a coding agent. Most providers also differ a lot in how they are configured, so the performance and quality you get varies considerably.
So when you reach a model through some other aggregator, you cannot be sure you are getting the best version of the model you asked for. To address that, we did three things:
- We tested a selected set of models and talked with their teams about how best to run them.
- We then worked with several providers to make sure those models are served correctly.
- Finally we benchmarked the model-and-provider combinations and compiled a list we are willing to recommend.
Qialike Bee is an AI gateway that gives you access to those models.
How it works
Bee works like any other provider in qialike.
- Sign in to the qialike console, add your billing details, and create an API key.
- Point qialike at Bee. In the TUI, run
/models, choose＋ Add provider→＋ Add a custom provider, and fill in:- route id:
qialike - base URL:
https://qialike.com/bee/v1 - API key: the key you just created
- model ids: any ids from
https://qialike.com/bee/v1/models
- route id:
- Run
/modelsagain and pick the model you want. The selection applies to the live session and persists as the default.
If you prefer environment configuration, export the key instead — qialike resolves a route’s key from the environment before the credential store:
export QIALIKE_API_KEY=qk-...qialikeYou are charged per request, and you can add credits to your account at any time.
Endpoints
Every model is reachable through one of the protocol routes below. Pick the route that matches your client; the gateway routes each model family to the right upstream provider.
| Protocol | Endpoint | Authentication |
|---|---|---|
| OpenAI Chat Completions | https://qialike.com/bee/v1/chat/completions | Authorization: Bearer <key> |
| OpenAI Responses | https://qialike.com/bee/v1/responses | Authorization: Bearer <key> |
| Anthropic Messages | https://qialike.com/bee/v1/messages | x-api-key: <key> |
| Google Generative Language | https://qialike.com/bee/v1/models/<model-id>:generateContent (:streamGenerateContent?alt=sse) | x-goog-api-key: <key> |
| Model catalog | GET https://qialike.com/bee/v1/models | Authorization: Bearer <key> |
Streaming and non-streaming. Set the protocol’s normal flag — "stream": true for
chat/completions, responses and messages, or :streamGenerateContent?alt=sse for Google — and Bee
passes server-sent events through as they arrive. Omit it and you get one JSON response body. The
same model and key work either way.
Rate limits. Requests are limited per API key and per model, and some providers add token-rate
shaping; the free models have their own trial limits. Exceeding a limit returns HTTP 429 with a
Retry-After header and a JSON error body naming the limit (RateLimitError,
FreeUsageLimitError, MonthlyLimitError, UserLimitError, …). The public edge also applies a
per-IP limit (20 requests/second, burst 40) before requests reach the gateway. Usage is metered per
request against your balance, and the console shows your usage history.
The model id you configure in qialike uses the format qialike/<model-id>.
For example, for DeepSeek V4 Flash you would use qialike/deepseek-v4-flash.
Pricing
Pay-as-you-go. Prices are per 1M tokens, in USD. Where a model charges differently above a
long-context threshold, both tiers are shown as ≤ threshold / > threshold. Models listed on one
row share the same price.
| Model | Input | Output | Cached Read | Cached Write |
|---|---|---|---|---|
| Big Pickle, DeepSeek V4 Flash Free, MiMo-V2.5 Free, North Mini Code Free, Nemotron 3 Ultra Free | Free | Free | Free | - |
| MiniMax M3, MiniMax M2.7, MiniMax M2.5 | $0.30 | $1.20 | $0.06 | - |
| GLM 5.2, GLM 5.1 | $1.40 | $4.40 | $0.26 | - |
| GLM 5 | $1.00 | $3.20 | $0.20 | - |
| Kimi K2.7 Code | $0.95 | $4.00 | $0.19 | - |
| Kimi K2.6 | $0.95 | $4.00 | $0.16 | - |
| Kimi K2.5 | $0.60 | $3.00 | $0.10 | - |
| Qwen3.7 Max | $2.50 | $7.50 | $0.50 | $3.125 |
| Qwen3.7 Plus | $0.40 | $1.60 | $0.04 | $0.50 |
| Qwen3.6 Plus | $0.50 | $3.00 | $0.05 | $0.625 |
| Qwen3.5 Plus | $0.20 | $1.20 | $0.02 | $0.25 |
| DeepSeek V4 Pro | $1.74 | $3.48 | $0.145 | - |
| DeepSeek V4 Flash | $0.14 | $0.28 | $0.028 | - |
| Grok 4.5 (≤ 200K / > 200K) | $2.00 / $4.00 | $6.00 / $12.00 | $0.50 / $1.00 | - |
| Grok Build 0.1 | $1.00 | $2.00 | $0.20 | - |
| Claude Fable 5 | $10.00 | $50.00 | $1.00 | $12.50 |
| Claude Opus 4.8, 4.7, 4.6, 4.5 | $5.00 | $25.00 | $0.50 | $6.25 |
| Claude Sonnet 5 | $2.00 | $10.00 | $0.20 | $2.50 |
| Claude Sonnet 4.6 | $3.00 | $15.00 | $0.30 | $3.75 |
| Claude Sonnet 4.5 (≤ 200K / > 200K) | $3.00 / $6.00 | $15.00 / $22.50 | $0.30 / $0.60 | $3.75 / $7.50 |
| Claude Haiku 4.5 | $1.00 | $5.00 | $0.10 | $1.25 |
| Gemini 3.5 Flash | $1.50 | $9.00 | $0.15 | - |
| Gemini 3.1 Pro (≤ 200K / > 200K) | $2.00 / $4.00 | $12.00 / $18.00 | $0.20 / $0.40 | - |
| Gemini 3 Flash | $0.50 | $3.00 | $0.05 | - |
| GPT 5.6 Sol (≤ 272K / > 272K) | $5.00 / $10.00 | $30.00 / $45.00 | $0.50 / $1.00 | $6.25 / $12.50 |
| GPT 5.6 Terra (≤ 272K / > 272K) | $2.50 / $5.00 | $15.00 / $22.50 | $0.25 / $0.50 | $3.125 / $6.25 |
| GPT 5.6 Luna (≤ 272K / > 272K) | $1.00 / $2.00 | $6.00 / $9.00 | $0.10 / $0.20 | $1.25 / $2.50 |
| GPT 5.5 (≤ 272K / > 272K) | $5.00 / $10.00 | $30.00 / $45.00 | $0.50 / $1.00 | - |
| GPT 5.5 Pro | $30.00 | $180.00 | $30.00 | - |
| GPT 5.4 (≤ 272K / > 272K) | $2.50 / $5.00 | $15.00 / $22.50 | $0.25 / $0.50 | - |
| GPT 5.4 Pro | $30.00 | $180.00 | $30.00 | - |
| GPT 5.4 Mini | $0.75 | $4.50 | $0.075 | - |
| GPT 5.4 Nano | $0.20 | $1.25 | $0.02 | - |
| GPT 5.3 Codex, GPT 5.3 Codex Spark | $1.75 | $14.00 | $0.175 | - |
| GPT 5.2, GPT 5.2 Codex | $1.75 | $14.00 | $0.175 | - |
| GPT 5.1, GPT 5.1 Codex, GPT 5 Codex, GPT 5 | $1.07 | $8.50 | $0.107 | - |
| GPT 5.1 Codex Max | $1.25 | $10.00 | $0.125 | - |
| GPT 5.1 Codex Mini | $0.25 | $2.00 | $0.025 | - |
| GPT 5 Nano | $0.05 | $0.40 | $0.005 | - |
Auto-reload
If your balance goes below $5, Bee automatically reloads $20. You can change the amount, or disable auto-reload entirely, in the console.
Monthly limits
You can set a monthly usage limit for the whole workspace and for each team member. A $20 monthly limit means Bee will not spend more than $20 that month — but if auto-reload is enabled, your card can still be charged more than $20 when the balance drops below $5.
Contact us if you have any questions.
Privacy
All our models are hosted in China. Our providers follow a zero-retention policy and do not use your data for model training, with these exceptions:
- Free models (Big Pickle, DeepSeek V4 Flash Free, MiMo-V2.5 Free, North Mini Code Free, Nemotron 3 Ultra Free): during their free period, collected data may be retained and used to improve the model. Do not submit personal or confidential data.
- NVIDIA free endpoints (Nemotron 3 Ultra Free): trial use only. Sessions are logged for security and product improvement, not linked to your identity. See NVIDIA’s API Trial Terms.
- OpenAI APIs: requests are retained for 30 days per OpenAI’s data policies.
- Anthropic APIs: requests are retained for 30 days per Anthropic’s data policies.
- Cohere-served free routes: see Cohere’s Terms and Privacy Policy.
Teams
Bee works well for teams: invite teammates, assign roles, and curate the models your workspace can use. Workspace management is free for teams during the beta; we will share pricing details before that changes.
Roles
- Admin — manage models, members, API keys and billing.
- Member — manage only their own API keys.
Admins can set monthly spending limits per member to keep costs under control.
Model access
Admins can enable or disable specific models for the workspace; requests to a disabled model return an error. This is useful for keeping your team off a model whose free tier retains data.
Bring your own key
You can use your own OpenAI or Anthropic API key while still accessing the other models in Bee. Those tokens are billed directly by the provider, not by Bee — useful when your organization already has a key.
Goals
We created Bee to:
- Benchmark the best models and providers for coding agents.
- Give you access to the highest quality options without downgrading performance or routing to cheaper providers.
- Pass along price drops by selling at cost, with the only markup covering processing fees.
- Have no lock-in: use Bee with any coding agent, and any other provider with qialike.