AI Chat: Allow some premium models to be used by non-premium users

[petemill]

The models will have stricter limits from the server-side and have special labelling when those limits are reached on the client side.

[petemill]

I'm not sure we yet have the exact UI designed for the rate limiting error on freemium models. I imagine it's similar to this, except the action will be something like "switch to basic model" @mattmcalister @aguscruiz?

[aguscruiz]

Yeah I agree, instead of "retry" we could use that action.

[aguscruiz]

Just remembered we actually have that full design here:
https://www.figma.com/file/m0Gdbf0wtqyfEFGm32VLLc/Leo?type=design&node-id=950-6301&mode=design

[kjozwiak]

The above requires 1.62.147 or higher for 1.62.x verification 👍

[kjozwiak]

@SergeyZhukovsky looks like the above is labelled as OS/Android. Double checking if the above can even be checked on Android yet? Maybe we can quickly check the above via 1.63.x for Android to make sure it's also working?

[SergeyZhukovsky]

@kjozwiak yes it's applied for Android and could be checked same way as on desktop using Android nightly build.

[kjozwiak]

@kjozwiak yes it's applied for Android and could be checked same way as on desktop using Android nightly build.

Thanks @SergeyZhukovsky. @hffvld @Uni-verse when verifying this one, lets use 1.64.x and add the results/verification notes within this issue 👍

[MadhaviSeelam]

Verification PASSED using

Brave | 1.62.150 Chromium: 121.0.6167.85 (Official Build) (64-bit)
-- | --
Revision | c01f5cde4d5d5e41134c0ccdb1868acd80a64a34
OS | Windows 11 Version 22H2 (Build 22621.3007)

Case 1: Select AI model selection from brave://settings/leo-assistant

Mixtral - PASSED

1. installed 1.62.150
2. launched Brave (without company VPN & without staging flags)
3. visited wired.com in a new tab
4. clicked sidebar on the toolbar
5. clicked Leo on the sidebar
6. clicked Summarize button
7. clicked Accept and begin button
8. verified Mixtral is a default
9. loaded an article https://www.wired.com/review/oneplus-12/
10. clicked on Summarize this page
11. clicked on Suggest questions...

Confirmed non-premium users can use premium model Mixtral. After 3 attempts, got rate limiting dialog.

example	example	example

Claude Instant by Anthropic - PASSED

1. new profile
2. launched Brave (without company VPN & without staging flags)
3. opened brave://settings/leo-assistant in a new tab
4. selected Claude Instant from `Default model for new conversations dropdown
5. visited wired.com in a new tab
6. clicked sidebar on the toolbar
7. clicked Leo on the sidebar
8. clicked Summarize this page
9. clicked Accept and begin button
10. verified the page is summarized as expected
11. clicked on a link in the article (https://www.wired.com/story/chinese-startup-01-ai-is-winning-the-open-source-ai-race/)
12. clicked on Summarize this page
13. repeated steps 10 & 11

Confirmed non-premium users can use premium model Claude Instant by Anthropic.

Note:
Waited about 2 hours after I got initial rate limit model with Mixtral
Response rate limit reached dialog is shown after 7 responses.

example	example	example	example	example	example	example

Llama 2 13B by Meta - PASSED

1. new profile
2. launched Brave
3. opened brave://settings/leo-assistant and selected Llama 2 13b
4. visited https://techcrunch.com/ in a new tab
5. clicked sidebar on the toolbar
6. clicked Leo on the sidebar
7. clicked Summarize this page
8. clicked Accept and begin button
9. verified page was summarized
10. clicked an article link https://techcrunch.com/2024/01/23/iphone-users-should-turn-on-apples-stolen-device-protection-feature/
11. clicked on Summarize this page
12. repeated 10 & 11 steps about 5 times.

Confirmed Responses returned as expected and rate limit dialog is not shown after repeating about 5 times.

example	example	example	example	example

Case 2: Select AI model from ... in the Leo panel -PASSED

1. new profile
2. launched Brave
3. visited https://techcrunch.com/ in a new tab
4. clicked sidebar on the toolbar
5. clicked Leo on the sidebar
6. clicked Summarize this page
7. clicked Accept and begin button
8. verified page was summarized
9. summarized the page successfully
10. confirmed got rate limit dialog after 5 responses
11. selected Llama 2 13b from ... from Leo pane
12. verified Leo responses and didn't get rate limiting model for free model
13. selected Claude Instant from ... from Leo pane
14. verified Leo responses and got rate limiting dialog for Claude Instant model

example	example	example	example	example	example	example	example	example	example

[hffvld]

Based on #34910 (comment) verified on Pixel 7 using version(s):

Device/OS: Pixel 7 / panther_beta-user 14 AP11.231215.007 release-keys
Brave build: 1.64.20 
Chromium: 121.0.6167.85 (Official Build) canary (64-bit) 

Filed follow-up issue #35560

Mixtral

STEPS:

1. Install and launch Brave 1.64.20
2. Visit any website, for example, https://chromeunboxed.com
3. Three-dot menu > Leo
4. Tap Summarize this page > Accept and begin > Verify
5. Tap the Three-dot menu in Leo > Verify
6. Go to any article on https://chromeunboxed.com
7. Three-dot menu > Leo > Tap Summarize this page
8. Tap on Suggest questions... > Verify
9. Tap suggested questions or make a new query > Verify

ACTUAL RESULTS:

• Verified that summarize this page is working
• Verified that Mixtral is set by default
• Verify that Suggest questions... is working
• Verified that non-premium users can use the premium model Mixtral. After 3 attempts, got the rate-limiting dialog Response rate limit reached

---

1	2	3

Claude Instant

STEPS:

1. Install and launch Brave 1.64.20
2. Visit any website, for example, https://chromeunboxed.com
3. Three-dot menu > Leo
4. Tap Summarize this page > Accept and begin > Verify
5. Three-dot menu on Leo page > Choose Claude Instant
6. Tap Summarize this page > Verify
7. Go to any article on https://chromeunboxed.com
8. Three-dot menu > Leo > Tap Summarize this page
9. Tap a link in the article > Three-dot menu > Leo > Tap Summarize this page
10. Repeat Steps 9 and 10 a few times

ACTUAL RESULTS:

• Verified that Mixtral is set by default
• Verified that summarize the page is working when switched to Claude Instant language model
• Verified that non-premium users can use the premium model Claude Instant. After 7 attempts, got the rate-limiting dialog Response rate limit reached

---

1	2	3

Llama 2 13B

STEPS:

1. Install and launch Brave 1.64.20
2. Visit any website, for example, https://chromeunboxed.com
3. Three-dot menu > Leo
4. Tap Summarize this page > Accept and begin > Verify
5. Three-dot menu on Leo page > Choose Llama 2 13B
6. Tap Summarize this page > Verify
7. Go to any article on https://chromeunboxed.com
8. Three-dot menu > Leo > Tap Summarize this page
9. Tap a link in the article > Three-dot menu > Leo > Tap Summarize this page
10. Repeat Steps 9 and 10 ~5 times

ACTUAL RESULTS:

• Verified that Mixtral is set by default
• Verified that summarize the page is working when switched to Claude Instant language model
• Verified that responses returned as expected, and rate limit dialog is not shown after repeating about 5 times

---

1	2	3	4

[kjozwiak]

Removing the above from #35625 as Leo is targeting 1.63.x for Android.