Back to all plans
星火 API
Spark X2
iFlytek Spark open API—Spark X2/Ultra/Pro language models and speech stack with per-token or volume-pack billing
Token API
SubscriptionToken API
Core models
Spark X2Spark X2 FlashSpark UltraSpark ProSpark Pro-128KSpark LiteSpark Max中文识别大模型多语种识别大模型超拟人语音合成一句话复刻
Spark X2
Deep-reasoning flagship—¥2/1M tokens on X2/X1.5 tab (page also shows ¥2–3 band).
Spark X2 Flash
Value deep-reasoning—¥1–2/1M list; X2-Flash lite 100M tokens ¥200 (¥2/1M), tier-4 down to ~¥1/1M.
Spark Ultra
Flagship with search and FunctionCall—¥0.8/1M tokens list.
Spark Pro
Strong performance with 128K context and search—¥5/1M tokens list.
Spark Pro-128K
Pro 128K long-context line—separate pricing tab with Pro-style volume packs.
See the official site for more models
Additional core model names still appear above, with full details on the latest official page.
Plan details
Spark Lite
FreeInput
¥0
Output
Simple chat · fast response
Usage
Spark Lite at ¥0/1M tokens list—for simple generation and general chat—the lowest-cost Spark API entry.
Models
Route higher-quality tasks to Ultra, Pro, or Spark X2 lines.
Highlights
Free packs: 200K tokens personal / 1M enterprise on X2-Flash and other tabs.
Same name inside Coding Plan uses request-based subscription—not this open API path.
Same name inside Coding Plan uses request-based subscription—not this open API path.
Best for
Light chat, trials, and cost-sensitive scenarios
Spark Ultra
Flagship valueRecommendedInput
¥0.8
Output
per 1M tokens · search & FunctionCall
Usage
Spark Ultra at ¥0.8/1M tokens list—full features with search and FunctionCall—a strong production default.
Models
Escalate complex reasoning to Spark X2; keep Lite for ultra-light tasks.
Highlights
See Ultra tab for volume packs and concurrency specs.
Merged billing—no separate input/output unit rates published.
Merged billing—no separate input/output unit rates published.
Best for
Production chat, knowledge assistants, and FunctionCall apps
Spark X2 Flash
Deep reasoningInput
¥1
Output
List ¥1–2/1M · packs from ~¥10/1M
Usage
Spark X2 Flash list ¥1–2/1M with stronger agents and code—X2-Flash lite pack 100M tokens ¥200 (¥2/1M), down to ~¥1/1M at volume.
Models
For agents, code, and fast/slow thinking mixes—evaluate volume packs at steady volume.
Highlights
Free packs: 200K personal / 1M enterprise tokens.
More value-oriented deep reasoning versus Spark X2 flagship.
More value-oriented deep reasoning versus Spark X2 flagship.
Best for
Agents, code, and deep-reasoning workloads
Spark Pro
PerformanceInput
¥5
Output
per 1M tokens · 128K context
Usage
Spark Pro at ¥5/1M tokens list—top performance with 128K context and search for high-quality complex tasks.
Models
Pro 128K tab offers long-context packs; see Max tab for batch inference.
Highlights
Better as key-path escalation than default routing for all traffic.
Merged billing—page does not split input/output.
Merged billing—page does not split input/output.
Best for
High-quality generation, long context, and complex business flows
Notes
- X2-Flash pack examples: lite 100M tokens ¥200 (¥2/1M), tier-4 125B tokens ¥125,000 (¥1/1M)—confirm live on X2-Flash tab.
- Tabs also show pay-as-you-go list rates (e.g. Ultra ¥0.8, Pro ¥5, Lite ¥0/1M) alongside volume packs.
- Speech models (ASR/TTS/clone) use different billing units from text tokens—estimate on console for your product.
- Hosted open models on MaaS plaza belong on third-party API pages; Coding Plan bundles are in plans.json.
Supported coding tools
OpenAI-compatible APIWebSocket APIBatch inferenceSparkDesk
Pricing and model data sourced from official vendor websites
FAQ
General6 条