# \[Solved\] Add Claude 3 models

**URL:** <https://forum.cursor.com/t/solved-add-claude-3-models/3214>\
**Category:** Feature Requests\
**Created:** [March 5, 2024, 11:59am UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214 "2024-03-05T11:59:00Z")\
**Posts on this page:** 20\
**Page:** 1

<div class="post-metadata">

**Author:** ![mejiasdev](https://avatars.discourse-cdn.com/v4/letter/m/59ef9b/32.png) [@mejiasdev](https://forum.cursor.com/u/mejiasdev)\
**Post date:** [March 5, 2024, 11:59am UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/1 "2024-03-05T11:59:00Z")

</div>

**UPDATE: [Claude support](https://changelog.cursor.sh/) is shipped**

I was wondering if Cursor has any plans to introduce new models, such a new

- Claude 3 Haiku
- Claude 3 Sonnet
- Claude 3 Opus

they all have 200,000 tokens context and good benchmarks compared to GPT-4

---

<div class="post-metadata">

**Author:** ![fedeandri](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/fedeandri/32/1403_2.png) [@fedeandri](https://forum.cursor.com/u/fedeandri)\
**Post date:** [March 5, 2024, 12:18pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/2 "2024-03-05T12:18:42Z")

</div>

Also interested in Claude 3

---

<div class="post-metadata">

**Author:** ![Bloodhawk](https://avatars.discourse-cdn.com/v4/letter/b/22d042/32.png) [@Bloodhawk](https://forum.cursor.com/u/Bloodhawk)\
**Post date:** [March 5, 2024, 12:42pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/3 "2024-03-05T12:42:43Z")

</div>

Also interested. Claude 3 Opus especially.

---

<div class="post-metadata">

**Author:** ![deanrie](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/deanrie/32/1263_2.png) [@deanrie](https://forum.cursor.com/u/deanrie)\
**Post date:** [March 5, 2024, 2:49pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/4 "2024-03-05T14:49:25Z")

</div>

It will be more expensive than GPT-4.

---

<div class="post-metadata">

**Author:** ![odysseus0z](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/odysseus0z/32/122_2.png) [@odysseus0z](https://forum.cursor.com/u/odysseus0z)\
**Post date:** [March 5, 2024, 8:10pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/5 "2024-03-05T20:10:40Z")

</div>

Hearing great things about Claude 3 Opus. Would love to be able to use it

---

<div class="post-metadata">

**Author:** ![Mushin](https://avatars.discourse-cdn.com/v4/letter/m/df788c/32.png) [@Mushin](https://forum.cursor.com/u/Mushin)\
**Post date:** [March 5, 2024, 8:27pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/6 "2024-03-05T20:27:18Z")

</div>

Hopefully very soon.

---

<div class="post-metadata">

**Author:** ![Dreams](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/dreams/32/8283_2.png) [@Dreams](https://forum.cursor.com/u/Dreams)\
**Post date:** [March 5, 2024, 8:52pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/7 "2024-03-05T20:52:52Z")

</div>

Use Openrouter.ai for now

---

<div class="post-metadata">

**Author:** ![dioro](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/dioro/32/847_2.png) [@dioro](https://forum.cursor.com/u/dioro)\
**Post date:** [March 5, 2024, 9:17pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/8 "2024-03-05T21:17:33Z")

</div>

I highly doubt we will get Opus due to the cost.

- GPT-4-Turbo is **$0.01** per 1k input tokens and **$0.03** per 1k output tokens
- Cloude 3 Opus is **$0.015** per 1k input tokens and **$0.075** per 1k output tokens

So 50% increase in input cost and 150% increase in output cost. If we ever get it I would guess it would be either with a new higher plan and/or less than 10k context like current GPT-4

I am much more exited about their upcoming Haiku model which would be $0.25 per **1 MILION** input token and $1.25 per **1 MILION** output tokens. i.e. $0.00025 per 1k input and $0.00125 per 1k output tokens.

According to their benchmarks, it’s supposed to be better than GPT-4 at coding. Combine that with 200k context would be insane.

---

<div class="post-metadata">

**Author:** ![jaredcodling](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/jaredcodling/32/1409_2.png) [@jaredcodling](https://forum.cursor.com/u/jaredcodling)\
**Post date:** [March 5, 2024, 10:33pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/9 "2024-03-05T22:33:07Z")

</div>

For those of us that make money from our projects (not putting down those that don’t yet, just talking justification), it’s 100% worth it to pay the extra for Claude.

It has a WAY better understanding of Next JS 13/14 and the app router for example. The hours it would save make it worth every penny.

I almost have to stop using cursor over this issue, and I love cursor, it’s a game changer; but right now I’m copy/pasting from Claude more than using cursor.

---

<div class="post-metadata">

**Author:** ![s1spam](https://avatars.discourse-cdn.com/v4/letter/s/53a042/32.png) [@s1spam](https://forum.cursor.com/u/s1spam)\
**Post date:** [March 6, 2024, 12:30am UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/10 "2024-03-06T00:30:58Z")

</div>

Agreed, I’m already using my own API key to utilise gpt-4-0125-preview for updated training on certain docs and is paid off on the company card. Would love to see Claude model options regardless of the price. However given Cursor is funded by OpenAI, not sure if there are any agreements to only offer OpenAI services?

---

<div class="post-metadata">

**Author:** ![dioro](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/dioro/32/847_2.png) [@dioro](https://forum.cursor.com/u/dioro)\
**Post date:** [March 6, 2024, 1:05am UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/11 "2024-03-06T01:05:45Z")

</div>

Apparently you can already do that [Support local LLM's - #17 by deanrie](https://forum.cursor.com/t/support-local-llms/1099/17)

---

<div class="post-metadata">

**Author:** ![bradhilton](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/bradhilton/32/1410_2.png) [@bradhilton](https://forum.cursor.com/u/bradhilton)\
**Post date:** [March 6, 2024, 1:33am UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/12 "2024-03-06T01:33:21Z")

</div>

Would also love to see support to use Claude models with our own API key.

---

<div class="post-metadata">

**Author:** ![tangjun](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/tangjun/32/16356_2.png) [@tangjun](https://forum.cursor.com/u/tangjun)\
**Post date:** [March 6, 2024, 1:47am UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/13 "2024-03-06T01:47:07Z")

</div>

Because the Claude API is expensive, so I guess, if Claude is supported, it will be a new plan with higher price.

---

<div class="post-metadata">

**Author:** ![Niq](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/niq/32/4950_2.png) [@Niq](https://forum.cursor.com/u/Niq)\
**Post date:** [March 6, 2024, 1:50am UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/14 "2024-03-06T01:50:01Z")

</div>

Would be great if we had the option to add and use Claude with our own api key.

---

<div class="post-metadata">

**Author:** ![lukaemon](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/lukaemon/32/114304_2.png) [@lukaemon](https://forum.cursor.com/u/lukaemon)\
**Post date:** [March 6, 2024, 6:05am UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/15 "2024-03-06T06:05:07Z")

</div>

API key to cover claude 3 would be great. Some questions worth more resources to get it right.

---

<div class="post-metadata">

**Author:** ![codexplorer](https://avatars.discourse-cdn.com/v4/letter/c/45deac/32.png) [@codexplorer](https://forum.cursor.com/u/codexplorer)\
**Post date:** [March 6, 2024, 12:06pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/16 "2024-03-06T12:06:14Z")

</div>

very much interested in Claude 3 Opus model, please add it in cursor.

---

<div class="post-metadata">

**Author:** ![Taidan](https://avatars.discourse-cdn.com/v4/letter/t/5f8ce5/32.png) [@Taidan](https://forum.cursor.com/u/Taidan)\
**Post date:** [March 6, 2024, 9:36pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/17 "2024-03-06T21:36:44Z")

</div>

@Jakob is cursor in any way tied to only using openai models?

---

<div class="post-metadata">

**Author:** ![Jakob](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/jakob/32/93_2.png) [@Jakob](https://forum.cursor.com/u/Jakob)\
**Post date:** [March 6, 2024, 9:53pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/18 "2024-03-06T21:53:44Z")

</div>

@Taidan No, as far as I know.

---

<div class="post-metadata">

**Author:** ![LionSR](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/lionsr/32/27174_2.png) [@LionSR](https://forum.cursor.com/u/LionSR)\
**Post date:** [March 8, 2024, 8:28am UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/19 "2024-03-08T08:28:45Z")

</div>

Yes, with the api from openrouter.ai, Claude 3 can be used; however, the image inputs seem to be not functioning properly. It is very expensive too.

---

<div class="post-metadata">

**Author:** ![punkouter23](https://avatars.discourse-cdn.com/v4/letter/p/ee59a6/32.png) [@punkouter23](https://forum.cursor.com/u/punkouter23)\
**Post date:** [March 8, 2024, 2:43pm UTC](https://forum.cursor.com/t/solved-add-claude-3-models/3214/20 "2024-03-08T14:43:56Z")

</div>

Has anyone actually compared claude to chatgpt head to head on a project? I am starting to look into that today for my next mini project… and gemini too

[Next page](https://forum.cursor.com/t/solved-add-claude-3-models/3214.md?page=2)
