Any plans to add Deepseek R1 model alongside V3?
5 Likes
Gentle reminder: readthe forum first, and the general zeitgeist of the room (internet) and youâll be much more Toolâd-Up to ask the Cursor community such profundities as:
1- given the chatter on HN re: deepseeek shall sink OAI and suchâŚ
2-Deepseek is much greater context and blah blah.
(i dont need to delineateâŚ)
however,
The true Elephant in the room (no, not you (yet, Postgres) â is:
How does deepseekâs context window and ability to be not_tied to else_model result in how Cursor (as a both Company and Technology) provide and adapt:
Will the value of Cursor IDE to the user be based on the Cursor [eco_system, model, provider, US/CN, etcâŚ
Where is Cursor in this Hiarchy of needs.
Now - given their immense (and quiet impact) on the greater SWE mindscape (do you for once think that EVERAY SINGLE red_team in the WORLD hasnt been hammering the â â â â out of Agenti Comput - and specifically Cursor on every desktopâŚ
(as they already had $$billion solutions (think ye olde IBM, CITRIX, SIEMENS, AS/400 (thanks Maxwell InQtel) (yeah yeah, we know the cabal⌠(palantir and splunk be damned, but we wont even start with DarkApp (everyone knows the âDarkWebâ â The DarkState is far more insidious (it built the DW to feed itself⌠(CodeLaundering)
Cursor IS the IDE step_stone â and its not about the model - its about the modality for users to utiliseâŚ
'25 is the literally first step of the coyboybootâs on the moon of mars via AI. YeeHaw
Iâve tested R1 and it is absolutely insane, we need this in 0.45.0
6 Likes
May you please share your process for setting it up, tests etc.
(The goal would be for [this and all future models] - to start building out what would be a âCommunity_Testsâ for setups and testing workflows.
â
Agentically grabbed and with .modelrules whereby .cursorrules grabs inputs from .modelrules and .testrules for the various workflows:
â>cursor . --model deepseekR1 --workspace [XYZ] --cursorrules [ai_stack_def] --testrules [full_test_details.rmd]
etcâŚ
Did you do the usual mickey mouse lightweight test or did you test it against a real word app / codebase and how did it perform?
Any idea if we can actually use it with open router DeepSeek R1 - API, Providers, Stats | OpenRouter