Deeprelay

Your favorite open source models at a better price

[Deeprelay @deeprelayai]

DeepSeek V4.1 Flash 1017 is live on Deeprelay — 55% below DeepSeek’s official API price, 1M context.

Per 1M tokens
Input $0.0675
Cached $0.0014
Output $0.27

OpenAI-compatible: change your base URL, keep your code.

[Deeprelay @deeprelayai]

@slyck_tv81342 @Satoshi87164293 Our default thinking is set to high

[Deeprelay @deeprelayai]

DeepSeek V4.1 Flash 1017 — 55% below DeepSeek’s official API price.

Per 1M tokens
Input $0.0675
Cached $0.0014
Output $0.27

One OpenAI-compatible key. Change your base URL, keep your code.

[Deeprelay @deeprelayai]

@HaneWari Messaged you. Could you let us know the model?

[Deeprelay @deeprelayai]

DeepSeek V4.1 Flash 1017 is live on Deeprelay — 55% below DeepSeek’s official API price, 1M context.

Per 1M tokens
Input $0.0675
Cached $0.0014
Output $0.27

OpenAI-compatible: change your base URL, keep your code.

[Deeprelay @deeprelayai]

@Pfar42396 Cope

[Deeprelay @deeprelayai]

@HarryStebbings @jasonlk 🤣

[Deeprelay @deeprelayai]

@ZiXuan58205 Deepseek makes 80-90%+ margin on official prices. Our partners who deploy for us are willing to take less

[Deeprelay @deeprelayai]

@n39r19 We serve official precision so only FP4 is some of the MOE. Rest is mostly FP8

[Deeprelay @deeprelayai]

Same DeepSeek V4.1 Flash 1017, 55% below the official API price.

One OpenAI-compatible key. Change your base URL, keep your code.

[Deeprelay @deeprelayai]

@catonooka This is a 1017 version issue when using very short prompts

[Deeprelay @deeprelayai]

@ItsmeAjayKV 👋

[Deeprelay @deeprelayai]

@ramendik You mean peak and off-peak pricing? Peak and off-peak pricing is not relevant to the subscription. We serve models mostly through our partners so will edit that to make it more clear

[Deeprelay @deeprelayai]

@ramendik No and no 🙃

[Deeprelay @deeprelayai]

@ramendik Yep 1017 update and we serve both at official precision. FP4 MOE + FP8 dense weights for 4.1 Flash and FP8 for 5.3 Flash

[Deeprelay @deeprelayai]

@haoguo123 Send us a message and we will look into it

[Deeprelay @deeprelayai]

New science benchmark weightings drag down V4.1 Flash's score but are not important for standard coding workflows

V4.1 Flash is still #1 in coding for its efficiency and the 🐳 will continue eating token volume

[AiBattle @AiBattle_]

The AA-Intelligence Index has been updated to V5

Rankings before:

1. Claude Opus 5.5 (Max) - 58
2. Claude Sonnet 5.5 (Max) - 56
3. Muse Wonder Full (Max)- 56
4. Kimi K3.1 Pro - 55
5. Claude Fable 5.1 (Max) - 53

Rankings now:

1. Claude Opus 5.5 (Max) - 58
2. GPT-6-Astra (Max) - 57
3. GPT-6.1-Sol (Max)- 55
4. Claude Sonnet 5.5 (Max) - 54
5. GPT-6.1-Terra (Max) - 51

[Deeprelay @deeprelayai]

@WarpSpeed17 DeepSeek makes a huge margin lol. We take less to give you a better price

[Deeprelay @deeprelayai]

@hybirdssss Have ZDR agreements on inputs and outputs with our upstreams. Cache hit depends on your traffic. Have customers running at 99%

[Deeprelay @deeprelayai]

@BlueGeorgill Yep