Official API Plan Comparison
Compare free quotas, top-up tiers, context windows, and rate limits across official DeepSeek, Qwen, Kimi, GLM, and StepFun APIs. Manually verified on Sep 20, 2026.
Per-model unit prices live on the model API price comparison.
DeepSeek
Official console| Plan | Billing | Free quota | Context | Rate limits | Notes | Source |
|---|---|---|---|---|---|---|
| Pay as you go | Cost = token usage × model price, deducted from topped-up or granted balance (granted balance first); no plan or top-up tiersPay as you go (unit prices on the Token API page) | — | 1M context; 384K max output (same for deepseek-flash and deepseek-v4-pro) | Account-level concurrency: deepseek-flash 2,500 / deepseek-v4-pro 500; higher limits via a support ticket at no extra cost | Peak hours (Beijing time Mon–Fri 9:00–12:00 and 14:00–18:00) cost more; off-peak prices are half of peak prices. | SourceSep 20, 2026 |
Alibaba Cloud Model Studio (Qwen)
Official console| Plan | Billing | Free quota | Context | Rate limits | Notes | Source |
|---|---|---|---|---|---|---|
| New-user free quota | Granted per model after activating Model Studio; covers real-time model calls only, not Batch, fine-tuning, deployment, or storage requestsFree | Independent quota per model, usually 1M tokens (shared input/output); valid 90 days from activation, model release, or approval; North China 2 (Beijing) region only | — | — | No real-name verification needed to claim; after exhaustion, unverified accounts are blocked while verified ones switch to pay-as-you-go; dated snapshot models have separate quotas; re-registering under the same verified entity does not reset quotas. | SourceSep 20, 2026 |
| Pay as you go and rate limits | Commercial models are billed by usage; billing and rate limits are independent, and topping up does not raise default limitsPay as you go (unit prices on the Token API page) | — | — | Rate limits apply per Alibaba Cloud account (RPM/TPM, merged across RAM users, workspaces, and API keys) and are independent per model; Batch calls bypass rate limits (queued); some models soft-limit by monthly spend tier | Higher limits require contacting sales; stable or latest versions get looser limits than dated snapshots. | SourceSep 20, 2026 |
Moonshot AI (Kimi)
Official console| Plan | Billing | Free quota | Context | Rate limits | Notes | Source |
|---|---|---|---|---|---|---|
| Tier 0 | Pay as you go; rate tiers upgrade automatically with cumulative top-up¥0 cumulative top-up (available at sign-up) | — | — | Concurrency 1 · RPM 3 · TPM 500,000 · TPD 1,500,000 · web-search QPS 1 | Vouchers do not count toward cumulative top-up; web-search QPS is independent of RPM/TPM/TPD and concurrency. | SourceSep 20, 2026 |
| Tier 1 | Pay as you go; rate tiers upgrade automatically with cumulative top-up¥50 cumulative top-up | — | — | Concurrency 15 · RPM 100 · TPM 2,000,000 · TPD unlimited · web-search QPS 3 | — | SourceSep 20, 2026 |
| Tier 2 | Pay as you go; rate tiers upgrade automatically with cumulative top-up¥100 cumulative top-up | — | — | Concurrency 40 · RPM 100 · TPM 3,000,000 · TPD unlimited · web-search QPS 5 | — | SourceSep 20, 2026 |
| Tier 3 | Pay as you go; rate tiers upgrade automatically with cumulative top-up¥500 cumulative top-up | — | — | Concurrency 50 · RPM 200 · TPM 3,000,000 · TPD unlimited · web-search QPS 10 | — | SourceSep 20, 2026 |
| Tier 4 | Pay as you go; rate tiers upgrade automatically with cumulative top-up¥5,000 cumulative top-up | — | — | Concurrency 60 · RPM 200 · TPM 4,000,000 · TPD unlimited · web-search QPS 20 | — | SourceSep 20, 2026 |
| Tier 5 | Pay as you go; rate tiers upgrade automatically with cumulative top-up¥20,000 cumulative top-up | — | — | Concurrency 100 · RPM 300 · TPM 5,000,000 · TPD unlimited · web-search QPS 50 | At cluster capacity limits, the platform may temporarily adjust rate limits. | SourceSep 20, 2026 |
Z.AI (GLM)
Official console| Plan | Billing | Free quota | Context | Rate limits | Notes | Source |
|---|---|---|---|---|---|---|
| Free models | Flash models marked free are callable at no cost; no plan purchase requiredFree | GLM-4.7-Flash, GLM-4-Flash-250414, GLM-Z1-Flash, GLM-4.6V-Flash, GLM-4V-Flash, GLM-4.1V-Thinking-Flash, CogView-3-Flash, CogVideoX-Flash, and similar models are free for input and output | GLM-4.7-Flash 200K; GLM-4-Flash-250414 and GLM-Z1-Flash 128K; GLM-4.1V-Thinking-Flash 64K; GLM-4V-Flash 4K | — | Cache storage is currently free for a limited time. | SourceSep 20, 2026 |
| Pay as you go | Text models are billed per token (CNY per million tokens) with input, output, and cache hits billed separately: cost = uncached input + cache hits + output + cache storage; images/videos per call, voice per 10K characters or minutePay as you go (unit prices on the Token API page) | — | — | — | Batch API models cost 50% of standard prices; some models tier input prices by per-request input length, per the official pricing table. | SourceSep 20, 2026 |
| Dedicated model instances | Billed by compute units occupied and days used, for stability and isolation needsCNY 100–200 per compute unit per day | — | — | — | Language models are mostly CNY 100 per compute unit per day, with GLM-4.6 (200K-fp8) and GLM-4.5 (128K-fp8) at 175; vision models 100 or 200; annual private-cloud packages also exist (GLM-4.5 CNY 1.1M/year, GLM-4.5-Air CNY 500K/year). | SourceSep 20, 2026 |
StepFun
Official console| Plan | Billing | Free quota | Context | Rate limits | Notes | Source |
|---|---|---|---|---|---|---|
| Step Plan activity credit | Step Plan time claimed during the activity: 15 days on sign-in to the open platform, plus 15 more days after your first API callFree to claim | 15 days on sign-in plus 15 days after the first API call; invites used to add 15 days per registered friend (up to 45 days combined), but the official page currently marks the invite reward as discontinued | — | — | Official API base is https://api.stepfun.com/v1; the page currently opens at most 100 new experience slots per day, and granting rules follow the activity page. | SourceSep 20, 2026 |
| Mini (monthly) | Monthly subscription; 400M Credits of monthly usage, entry tier¥49/month (currently marked sold out on the page) | — | — | — | Covers all flagship models with smart routing, multiple MCP tools, and multi-device sign-in; Studio adds an extra creative quota worth 40% of the plan. | SourceSep 20, 2026 |
| Plus (monthly) | Monthly subscription; 1,600M Credits of monthly usage, daily-driver tier¥99/month (currently available on the page) | — | — | — | Includes everything in Mini, plus priority API rate and priority support. | SourceSep 20, 2026 |
| Pro (monthly) | Monthly subscription; 8,000M Credits of monthly usage, professional tier¥199/month (currently marked sold out on the page) | — | — | — | Includes everything in Mini, plus priority API rate and priority support. | SourceSep 20, 2026 |
| Max (monthly) | Monthly subscription; 40,000M Credits of monthly usage, top tier¥699/month (currently marked sold out on the page) | — | — | — | Includes everything in Mini, plus priority API rate and priority support; the page says quarterly or annual billing saves up to CNY 1,722. | SourceSep 20, 2026 |