Xfyun
SG
Latency
-
Throughput
41tps
Uptime
Total Context
128K
Max Output
-
Status
Standby
Pricing
Input Price
Provider priced
/M tokens
Output Price
Provider priced
/M tokens
Cache Creation (5min)
-
/M tokens
Cache Read (5min)
-
/M tokens
X
Xfyun: Spark Pro-128K
xfyun/spark-pro-128k
Current versionSpark Pro-128KModestable
讯飞星火专业长上下文模型，适合文档阅读、知识问答和办公摘要
Modalities
Input
Provider priced per 1M
Output
Provider priced per 1M
context
128K
Weekly Tokens
Live catalog
Released
Live catalog
Overview for Spark Pro-128K
A compact summary of this model's best-fit workloads, routing status and production integration shape.
Architecture
Xfyun OpenAI-compatible route
Model design signal
Traffic signal
Live catalog
Recent public usage volume
Benchmark score
-
Composite coding and reasoning signal
Best for
long contexttext generation
Model tags
chatstreaminglong-context
Routing snapshot
Primary provider
Xfyun
Provider region
Provider policy
Current version
Spark Pro-128K
Routing health
Missing key
Missing XFYUN provider key
Xfyun
Fallback route
Context and limits
Total Context
128K
Max prompt
128K
Max Output
-
Pricing tier
x1.0 gateway rate
Pricing snapshot
Input PriceProvider priced
Output PriceProvider priced
Cache creation-
Cache read-
Integration checklist
✓OpenAI-compatible chat completions
✓Call with model ID xfyun/spark-pro-128k
✓Use base URL https://ethergate.cn/v1
✓Create a workspace-scoped API key
Capability profile
chatStreaminglong-context
Xfyun policies apply for retention, training and regional handling.
Parameter support
max_tokenstemperaturetop_pCheck long-context pricing once prompts exceed 128K.
Very large prompts can add latency; test production payloads before launch.
Providers for Spark Pro-128K
EtherGate routes requests to configured providers that can handle your prompt size and parameters, with fallbacks to maximize uptime. uptime.
Performance for Spark Pro-128K
Compare different providers across EtherGate
Throughput
30dXfyunAverage 40 tok/s
Latency
30dXfyunAverage 1.84 s
E2E Latency
30dXfyunAverage 12.54 s
Tool Call Error Rate
30dXfyunAverage 4.83 %
Structured Output Error Rate
30dXfyunAverage 1.38 %
Effective Pricing for Spark Pro-128K
Actual cost per million tokens across providers over the past hour
Weighted Average
Weighted Avg Input Price
Provider priced
Weighted Avg Output Price
Provider priced
| Provider | Input Price | Output Price | Cache Hit Rate |
|---|---|---|---|
| Xfyun | Provider priced | Provider priced | 42% |
Input Price / 1M tokens
7dMay 12May 14May 16May 18
Output Price / 1M tokens
7dMay 12May 14May 16May 18
Benchmarks for Spark Pro-128K
Performance metrics and benchmarks
Ranking Distribution
1617
#1
1653
#2
1322
#3
1368
#4
Category Performance
Composite coding and reasoning signal
Score
-
Models Arena
3D
1297
53.0% Win · 170.0s Avg
ASCII
1196
50.0% Win · 152.0s Avg
Code Categories
1288
53.4% Win · 254.7s Avg
Data Visualization
1283
53.3% Win · 236.9s Avg
Game Development
1292
54.8% Win · 270.4s Avg
Image to HTML
1205
46.7% Win · 116.4s Avg
SVG
1239
55.2% Win · 133.4s Avg
UI Component
1303
54.5% Win · 252.7s Avg
Website
1266
51.8% Win · 306.2s Avg
Apps using Spark Pro-128K
Top public apps this month
Missing key: Missing XFYUN provider key
1
Hermes Agent
Open-source agent workspace for model routing
353B
Tokens
2
OpenClaw
AI coding agent connected to EtherGate
116B
Tokens
3
AA-LCR Benchmark
Applied research lab for frontier models
35.9B
Tokens
4
Claude Code Bridge
Developer tool using OpenAI-compatible APIs
26.8B
Tokens
5
Roo Code
VS Code extension for model-assisted coding
17.3B
Tokens
Monthly token usage
Apr 20May 1May 10May 18
Recent activity on Spark Pro-128K
Total usage per day on EtherGate
Prompt tokensReasoning tokensCompletion tokens
Uptime stats for Spark Pro-128K
Uptime stats for Spark Pro-128K across configured providers
Average uptime
Not configured
Sample code and API for Spark Pro-128K
EtherGate normalizes requests and responses across providers for you.
EtherGate supports reasoning-enabled models that can return step-by-step reasoning metadata. Use the reasoning parameter when a provider supports it, and preserve reasoning_details when continuing a conversation. Learn more about reasoning tokens.
Request
Use the OpenAI-compatible Chat Completions endpoint. Keep your SDK, change the base URL, API key and model ID.
OpenAI SDK
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.ETHERGATE_API_KEY,
baseURL: "https://ethergate.cn/v1",
});
const completion = await client.chat.completions.create({
model: "xfyun/spark-pro-128k",
messages: [{ role: "user", content: "Explain your best use case." }],
});Tool configuration examples
Use this model ID directly in agent tools that accept OpenAI-compatible endpoints.
QClaw macOS/Linux
curl -fsSL 'https://ethergate.cn/console/skills/ethergate-provider/install.sh' | bash -s -- --target qclaw --api-key '<your-ethergate-key>' --locale 'en'Required headers
Authorization
Bearer $ETHERGATE_API_KEYHeader
Content-Type: application/jsonBase URL
https://ethergate.cn/v1Optional headers
Optional attribution headers help you identify traffic by app or website when debugging usage.
HTTP-Refererhttps://your-app.example
X-TitleYour app name
Using third-party SDKs
openai-js
openai-python
Vercel AI SDK
LangChain