RelayRouterModels / Compare / kimi-k3 vs deepseek-v4-flash
kimi-k3 vs deepseek-v4-flash
Both models are available on RelayRouter through one base URL. This page shows which public groups serve each one, what the groups charge, and the protocol to call, so you can pick by price and integration effort. It does not rank quality.
kimi-k3
kimi-k3
- Provider
- Moonshot AI
- Modality
- text generation
- Public groups
- 1
- Cheapest group
- Kimi K3: ¥16.32 (about $2.40) per 1M input tokens, ¥81.6 (about $12.00) per 1M output tokens
- Protocols
- Anthropic Messages API
deepseek-v4-flash
deepseek-v4-flash
- Provider
- DeepSeek
- Modality
- text generation
- Public groups
- 1
- Cheapest group
- DeepSeek: ¥1.1 (about $0.162) per 1M input tokens, ¥4.4 (about $0.647) per 1M output tokens; off-peak; ×2 on weekdays 09:00 to 12:00 and 14:00 to 18:00 Beijing time
- Protocols
- Anthropic Messages API
Same request, two models
import anthropic
client = anthropic.Anthropic(
api_key="your-relayrouter-key",
base_url="https://relayrouter.io",
)
msg = client.messages.create(
model="kimi-k3",
max_tokens=1024,
messages=[{"role": "user", "content": "Hello!"}],
)
print(msg.content[0].text)import anthropic
client = anthropic.Anthropic(
api_key="your-relayrouter-key",
base_url="https://relayrouter.io",
)
msg = client.messages.create(
model="deepseek-v4-flash",
max_tokens=1024,
messages=[{"role": "user", "content": "Hello!"}],
)
print(msg.content[0].text)Questions
Is kimi-k3 or deepseek-v4-flash cheaper on RelayRouter?
Compare the cheapest group of each: kimi-k3 is served by "Kimi K3" (¥16.32 (about $2.40) per 1M input tokens, ¥81.6 (about $12.00) per 1M output tokens) and deepseek-v4-flash by "DeepSeek" (¥1.1 (about $0.162) per 1M input tokens, ¥4.4 (about $0.647) per 1M output tokens; off-peak; ×2 on weekdays 09:00 to 12:00 and 14:00 to 18:00 Beijing time). Settlement rates apply to the provider's standard usage, so a model with a higher official price can still cost more at a lower rate.
Can I switch between the two without changing code?
If both models sit in groups with the same protocol you only change the model name. If the protocols differ, point the matching SDK at https://relayrouter.io; the base URL and key handling stay the same.
Does this page compare output quality?
No. It compares availability, protocols and price on RelayRouter. Run your own evaluation prompts through both models; failed requests are not billed.
Rates are indicative; the console shows the authoritative rate at request time. See pricing and the full model list.