SparkCode Models 2026: Every Supported LLM Tested — Multi-model users who want low commitment

The 30-second summary

In-depth review

2 models listed; GPT-4o and Claude 3.5 Sonnet are the headline names, but the platform description mentions broader coverage including Gemini, Kimi, and GLM.

Model-by-Model Breakdown

Don’t let the short public model list fool you — SparkCode’s registration page reveals more. The core offering is straightforward: GPT-4o and Claude 3.5 Sonnet for standard relay usage, with access to Kimi, GLM, and Gemini listed as additional options.

GPT-4o

Standard OpenAI relay behavior. Context window hits 100k tokens, which matches the official GPT-4o API limit. Speed is average for a relay — expect 30-40 tokens/second during peak hours. The 94% uptime figure means you’ll occasionally hit a 5-minute outage window each day. Not terrible for a free-tier entry.

Claude 3.5 Sonnet

This is where SparkCode gets interesting. The Opus 4.6 pricing at ¥5 (in) / ¥25 (out) per million tokens is genuinely cheap — roughly 60% below Anthropic’s direct API pricing. The catch? Quality inconsistency. Some requests return sharp, coherent responses; others feel like the relay is compressing or truncating outputs. Test your prompts before committing to production use.

Gemini, Kimi, GLM

These are listed as supported but the platform data flags “Gemini resources frequently unavailable.” If you need Gemini for multimodal tasks, look elsewhere. Kimi and GLM are stable but slow — expect 15-20 tokens/second. They work for batch processing where latency doesn’t matter.

Pricing Table

The ¥1 minimum recharge is the standout feature here. You can test both GPT-4o and Claude 3.5 Sonnet for the price of a bus ticket. No promo code needed — the ¥2 free credit covers initial testing.

Pros & Cons

Pros

• ¥1 minimum recharge — lowest barrier to try in this review set
• Opus 4.6 pricing at ¥5(进)25(出) is genuinely competitive
• ¥2 free credit means zero-cost evaluation
• Supports 5+ model families from a single relay

Cons

• Gemini availability is unreliable — don’t depend on it
• Monthly subscription offers almost no savings over PAYG
• Quality inconsistency on Claude 3.5 Sonnet outputs
• Uptime at 94% is below the 97%+ standard for production relays

Verdict

SparkCode is a “try-before-you-buy” relay. The ¥1 minimum recharge and ¥2 free credit make it the cheapest way to test GPT-4o and Claude 3.5 Sonnet without a VPN. But the 94% uptime and quality inconsistency mean I wouldn’t run production workloads here.

Use it for: prototyping, personal projects, or as a backup relay when your primary provider has an outage. Skip it if you need reliable Gemini access or consistent Claude output quality. The pricing is good, but the reliability isn’t there yet.

FAQ

Q: Can I use SparkCode without registering a payment method? A: Yes. The ¥2 free credit on registration lets you test both GPT-4o and Claude 3.5 Sonnet without adding 支付宝 or 微信支付. You’ll need to add payment only when the free credit runs out.

Q: Does SparkCode support streaming responses? A: The platform data doesn’t specify streaming support explicitly. Based on the 100k token context window and standard relay behavior, streaming likely works but expect the same quality inconsistency noted for Claude 3.5 Sonnet.

Q: What happens when Gemini resources are unavailable? A: The API will return an error or timeout. SparkCode doesn’t auto-fallback to another model. You’ll need to handle this in your application logic — retry with a different model or queue the request for later.

Pricing breakdown

SparkCode offers competitive pricing for developers. Here's the breakdown:

Supported models

2 models across major vendors.

Frequently asked questions

Should you use SparkCode?

Multi-model users who want low commitment