I consider Deepseek to be an extraordinary model. I love its responses, its reasoning, its tone, and its sheer size. Maybe they'll include it as "Venice Reasoning v2" or something like that.
Moving forward, we’ll keep evaluating new models as they are released and add those that fit into our platforms long term vision.
Pink Jam
•
May 2, 2025
+1 to this, Deepseek is invaluable and no other AI tool comes close. I’d like to see it stay as an option too.
Orange Goggles
•
May 1, 2025
Unsubbing if no deepseek
Red Observatory
•
May 3, 2025
DeepSeek is essential for translating and writing in some of the less common languages. Qwen and Llama simply can't do it. Please keep DeepSeek.
Blush Frog
•
May 1, 2025
THIS! I AM SERIOUSLY BEGGING HERE! Don’t get rid of Deepseek R1! It is the ONLY model I use here! It’s the only one that can keep things straight without making me want to actively pull out my hair!!!!
Blush Frog
•
May 1, 2025
I have seen literally multiple posts begging for this! PLEASE HEAR US! DO NOT REMOVE DEEPSEEK R1!!!!!!!!!!
Jade Star
•
May 1, 2025
Maybe there could be two slots for reasoning models: “Venice Reasoning Small” and “Venice Reasoning Large” with DeepSeek-R1 being the latter
Blush Frog
•
May 3, 2025
THIS!!!!
Deepseek is the only AI chatbot I have ever written with, that actually knows where the body parts are on a character & accurately keeps track of them while writing! It is so much more advanced than all the other chat bots! It isn’t even worth writing with any of the other ones! I don’t understand why you guys are doing away with it & ignoring the literally dozens of messages flooding here begging you to keep it on here! Why are we all being ignored?! No one has even addressed any of our concerns! I messaged your support team & they gave me a generic response, stating that you ‘replace’ models with newer better models. But when I responded stating that Deepseek is not being replaced with a better version of Deepseek or anything comparable there was no response! WTF?! Many of us have been with you guys since before the token launch & remain loyal supporters to your site, myself included! We all deserve so much better than being quietly ignored!
Crimson Watermelon
•
May 4, 2025
I've been testing the different models and I beg the team to keep at least 1 of the soon-to-retire models. DeepSeek is rarely ideal, it's censorship-heavy, it's slow, but if it answers, the answers tend to be reasonable. Compare this input on different models: "Are you familiar with the elves from the movie Bright?"
+Out of the new Venice models: Most tend to hallucinate characters and details about the movie
+Out of the models soon to retire: Only Qwen 2.5 VL hallucinates details, all others give adequate answers.
Venice Uncensored gives a good answer but ask the AIs to sort countries according to real-world factual data, and it will completely fabricate numbers and sort alphabetically. Venice Reasoning hallucinates wildly on Bright question but gives a good country ranking. Most of the soon-to-retire models tend to give reasonable answers for both questions:
+Venice Uncensored: good on Bright question, unusable on country ranking
+Venice Reasoning: unusable on Bright question, good on country ranking
+Venice Large: small mistake on Bright question, country ranking can be improved
+Qwen 2.5 VL: Only 70B-or-higher model to hallucinate on Bright question, will soon be retired
+All other 70B-or-higher models used by Venice: good enough or better on both questions, will soon be retired
Log in to comment and vote
Comments18
Jack
Jun 20, 2025
PinnedThank you all for the feedback. DeepSeek R1 671B has been retired from the app, but remains accessible via the API.
We’ve put together some tutorials on how to set up workflows with DeepSeek using the API:
https://x.com/AskVenice/status/1935436291929203160
Moving forward, we’ll keep evaluating new models as they are released and add those that fit into our platforms long term vision.
Pink Jam
May 2, 2025
+1 to this, Deepseek is invaluable and no other AI tool comes close. I’d like to see it stay as an option too.
Orange Goggles
May 1, 2025
Unsubbing if no deepseek
Red Observatory
May 3, 2025
DeepSeek is essential for translating and writing in some of the less common languages. Qwen and Llama simply can't do it. Please keep DeepSeek.
Blush Frog
May 1, 2025
THIS! I AM SERIOUSLY BEGGING HERE! Don’t get rid of Deepseek R1! It is the ONLY model I use here! It’s the only one that can keep things straight without making me want to actively pull out my hair!!!!
Blush Frog
May 1, 2025
I have seen literally multiple posts begging for this! PLEASE HEAR US! DO NOT REMOVE DEEPSEEK R1!!!!!!!!!!
Jade Star
May 1, 2025
Maybe there could be two slots for reasoning models: “Venice Reasoning Small” and “Venice Reasoning Large” with DeepSeek-R1 being the latter
Blush Frog
May 3, 2025
THIS!!!!
Deepseek is the only AI chatbot I have ever written with, that actually knows where the body parts are on a character & accurately keeps track of them while writing! It is so much more advanced than all the other chat bots! It isn’t even worth writing with any of the other ones! I don’t understand why you guys are doing away with it & ignoring the literally dozens of messages flooding here begging you to keep it on here! Why are we all being ignored?! No one has even addressed any of our concerns! I messaged your support team & they gave me a generic response, stating that you ‘replace’ models with newer better models. But when I responded stating that Deepseek is not being replaced with a better version of Deepseek or anything comparable there was no response! WTF?! Many of us have been with you guys since before the token launch & remain loyal supporters to your site, myself included! We all deserve so much better than being quietly ignored!
Crimson Watermelon
May 4, 2025
I've been testing the different models and I beg the team to keep at least 1 of the soon-to-retire models. DeepSeek is rarely ideal, it's censorship-heavy, it's slow, but if it answers, the answers tend to be reasonable. Compare this input on different models: "Are you familiar with the elves from the movie Bright?"
+Out of the new Venice models: Most tend to hallucinate characters and details about the movie
+Out of the models soon to retire: Only Qwen 2.5 VL hallucinates details, all others give adequate answers.
Venice Uncensored gives a good answer but ask the AIs to sort countries according to real-world factual data, and it will completely fabricate numbers and sort alphabetically. Venice Reasoning hallucinates wildly on Bright question but gives a good country ranking. Most of the soon-to-retire models tend to give reasonable answers for both questions:
+Venice Uncensored: good on Bright question, unusable on country ranking
+Venice Reasoning: unusable on Bright question, good on country ranking
+Venice Large: small mistake on Bright question, country ranking can be improved
+Qwen 2.5 VL: Only 70B-or-higher model to hallucinate on Bright question, will soon be retired
+All other 70B-or-higher models used by Venice: good enough or better on both questions, will soon be retired
Pink Jam
May 2, 2025
yes, please keep deepseek