wtf is deepseek r1? v3?

R1 as a reasoning powerhouse || what's a reasoning model? || wtf is the difference?

"The study of reasoning model draws its inspiration from the observation that reasoning is one of the main intelligent activities of human beings, and of the ways by which human thinking comes from one idea to a related idea." – Science Direct

ty, smart person.

Deepseek-R1 is a reasoning model (unlike GPT-4o, Sonnet3.5, JLLM, etc. When it thinks about your prompt and instruction, this is its thought process (relative):

1. First, I should set the scene...

2. Next, sensory details...

3. I need to show their internal state…

4. Ending with a poignant image...

5. Let me check if I'm covering all elements...

[More information in this article: DeepSeek R1 vs V3 on datacamp.com]

The result:

• thinking output (because of CoT= Chain of Thinking, aka: "Okay, let me think about this...") ⇒ slower response time

• stricter adherence to API— you need to be specific about exactly WHAT and HOW you want it to act.

• very fussy about how you structure your prompt, what words you use etc.

• however, when it understands you, it cooks a full-course meal

Differences between R1 and V3?

i'll leave you with this from datacamp.com, link to article here

DeepSeek-V3 doesn’t have reasoning ability. As we mentioned, it works as a next-word predictor. This means that it can answer questions whose answers are encoded in the training data.

Because the amount of data used to train these models is so huge, it is capable of answering questions on almost any topic. Like other LLMs, it excels in natural-sounding conversation and creativity. It's the model we want for creating writing, content creation, or answering generic questions that are likely to have been solved already numerous times.

DeepSeek-R1, on the other hand, shines when it comes to complex problem-solving, logic, and step-by-step reasoning tasks. It's designed to tackle challenging queries that require thorough analysis and structured solutions. When faced with complex coding challenges or detailed logical puzzles, R1 is the tool to rely on.

source: datacamp.com

so now you know!

---

For more troubleshooting and setup tips (also how to set up Chutes), here is a very good rentry for you.

HOW TO USE DEEPSEEK MODELS AS A PROXY || JANITOR.AI SPECIFIC (paid + free)

1. Official DeepSeek API (paid)

1

Make an account on the Deepseek Platform

2

Generate a key

3

Add credits (top up your account)

4

Go back to Janitor

Go here:

CHOOSE Proxy, then paste the key here:

5

Add the model + proxy URL

R1: deepseek-reasoner

V3: deepseek-chat

Proxy URL: https://api.deepseek.com/v1/chat/completions

6

Save, refresh and done!

THE TEMP ISSUE (OFFICIAL API)

‼️ With this method, TEMP doesn't affect R1. find the official docs here

FOR V3: this is how temp works for this model (info can be found here)

IF set temp = 1, THEN actual temp = 0.3

1. OpenRouter method (free****‼️)

1

Make an account on Openrouter.ai

2

Generate a key (Menu > Keys > Create Key)

COPY THE KEY INTO YOUR NOTES || looks something like sk-v1-xxxxxxx

3

Go back to Janitor

Go here:

CHOOSE Proxy, then paste the key here:

4

Add the model and proxy URL

Model: deepseek/deepseek-r1:free

For V3: deepseek/deepseek-chat-v3-0324:free

Proxy URL: https://openrouter.ai/api/v1/chat/completions

5

Save, refresh and done!

****—previously, OpenRouter provided 200 free messages. Nowadays, you get 50.

to get more, you must add $10 in credits to get 1000 free messages.

if this doesn't work for you, refer to #ai-models's (janitor discord channel) rentry right here on how to set up Chutes, which is free (last edit: 14 Apr 25)