The Prince of All Saiyans
Formerly Jisoo Stan & @Twitter
tf is okayplayer forums
Colibreh @bnew explains difference between ChatGPT and Chinese AI Deepseek
- Thread starter Seoul Gleou
- Start date
1/64
@Kimi_Moonshot
Introducing Kimi k1.5 --- an o1-level multi-modal model
-Sota short-CoT performance, outperforming GPT-4o and Claude Sonnet 3.5 on
AIME, MATH-500, 
LiveCodeBench by a large margin (up to +550%)
-Long-CoT performance matches o1 across multiple modalities (
MathVista, AIME, Codeforces, etc)
Tech report: GitHub - MoonshotAI/Kimi-k1.5
Key ingredients of k1.5
-Long context scaling. Up to 128k tokens for RL generation. Efficient training with partial rollouts.
-Improved policy optimization: online mirror descent, sampling strategies, length penalty, and others.
-Multi modalities. Joint reasoning over text and vision.
2/64
@Cryp70
Chat replies are in English but the rest of your site is Chinese. Is there a full English version available?
3/64
@Kimi_Moonshot
Coming soon, stay tuned!
4/64
@AntDX316
Make an English version app for iOS and Android.
5/64
@winchest_stella
Just tried it. its impressive esp search and vision capabilities. Big congrats to the team!
6/64
@Cryp70
Impressive scores, nice benchmark for code. DeepSeek is my go to but I see it's time to give Kimi a test
7/64
@inikhil__
Seems like china is winning.
8/64
@Saboo_Shubham_
This is awesome progress. Keep it coming.
9/64
@McGee_noodle
🫡
10/64
@CallMeSam89
This is huge!
Please bench against r1 as well.
11/64
@mark_k
"Joint reasoning over text and vision."
OMG this is huge. I wonder if it could be extended to other modalities too, e.g. audio?
12/64
@4K_Everyday
What the
🫡
13/64
@jrabell0
Where is GPQA?
14/64
@itsPaulAi
Two Chinese o1 models released on the same day? It's speeding up!
15/64
@Grynn
Prices, param counts, open-source? open-weights?
16/64
@DrJimFan
Love it. Keep up the great work!
17/64
@senb0n22a
Very strong search capabilities. I forgot other AIs aren't allowed to search Twitter, otherwise it might have been the best search parser for now. Can process 100+ web pages in one query, compared to others capping at 25-50. Instruction following isn't as strong as Deepseek/Grok 2, but for web research, I could recommend this one.
18/64
@hasantoxr
Wow this is super cool
19/64
@iamfakhrealam
I have recently installed the @deepseek_ai application and found it to be exceptionally amazing…
Could you kindly provide me with the link to the @Kimi_ai_ application?
20/64
@nisten
wait wut, ok this i need to test
21/64
@CodeByPoonam
Wow another Chinese o1 models outperforming ChatGPT.
22/64
@amoussouvichris
Amazing results, your team did an amazing job !
23/64
@praveenjothi99
why the mobile verification? almost no one asks!
24/64
@rohanpaul_ai
Beautiful..
25/64
@acharya_aditya2
Will we get open weights ??
26/64
@SmartFlowAITeam
Great ！！！
27/64
@rezkhere
That's a powerful model
28/64
@SkyBlueHarbor
english please, i'm excited to try it out
29/64
@jseles11
this and a Mac mini is all you really need
30/64
@AILeaksAndNews
What a day for Chinese AI
31/64
@TechByMarkandey
Seems amazing can we connect.
I cannot dm you
32/64
@Cory29565470
Where is @GoogleAI ?
33/64
@SaquibOptimusAI
Oh, bro. Another one.
"Make SOTA AI Cheap Again".
Awesome.
34/64
@DuckWithCup
I tried Kimi before and it’s amazing. Thank you Team.
35/64
@daily_ai_takes
Great work! Exciting times ahead
36/64
@CJ_Wolff
Is there an API
37/64
@DhruvmehtaRps
Where are the other benchmarks?
38/64
@MuchMore2It
Can you add it to @OpenRouterAI?
39/64
@Pedram_virus
When will it be possible to log in with Google? And when will full support for the English language be available? Because it is still in Chinese.
40/64
@Maeelk
Do you plan to open source as @deepseek_ai did ?
@huggingface still has a few To available I guess.
41/64
@FoundTheCode
o1 models everywhere, we're soo back
42/64
@thecute_8
是国产AI官方账号，支持一波
43/64
@ArpanTripathi20
@untitled01ipynb “Mr president a second o1-level model has dropped” Sam A replaced on George Bush’s face
44/64
@wojtess
Where can I find weights?
45/64
@DavidSZDahan
for Kimi's team if this model will not become open source like this tweet or reply with a .
46/64
@JennyZhang6989
@Kimi_Moonshot Where can we use short CoT in Kimi?
47/64
@DavidSZDahan
Where we can use it ?
48/64
@txhno
it's christmas
49/64
@Ttkouhe
When release. I wanan try!!
50/64
@URUBONZ_
Is your google login coming anytime soon? I have been unable to get SMS to send a confirmation and Id love the try the new version
51/64
@realmrfakename
Cool! Any plans to open source?
52/64
@_HARVEY__DENT_
Good grief
53/64
@Angelov_Sta
Why o1 benchmarks are so low? In the deepseek r1 comparisons, o1 scores higher vs what shown here
54/64
@playfuldreamz
Read the room
55/64
@SenougaharA
Looks good tbh. Just bad timing maybe. Still all the best because it does look good
56/64
@rose567888
57/64
@FyruzOne
How does it do on gpqa diamond
58/64
@the__sonik
Why can't we sign up on the website using Google? Is access restricted only to people in China?
59/64
@dabobo0496
加油
60/64
@Jane1374555767
这条帖子下面应该有一条简体中文回复。
61/64
@SonyxEth
is there an english version
62/64
@TadiwaClyde
Open source?
63/64
@tenmillioncoins
can i download this on ollama search
64/64
@bruce_x_offi
Are you planning to open-source it?
@Kimi_Moonshot
Introducing Kimi k1.5 --- an o1-level multi-modal model-Sota short-CoT performance, outperforming GPT-4o and Claude Sonnet 3.5 on
AIME, MATH-500, LiveCodeBench by a large margin (up to +550%)-Long-CoT performance matches o1 across multiple modalities (
MathVista, AIME, Codeforces, etc)Tech report: GitHub - MoonshotAI/Kimi-k1.5
Key ingredients of k1.5
-Long context scaling. Up to 128k tokens for RL generation. Efficient training with partial rollouts.
-Improved policy optimization: online mirror descent, sampling strategies, length penalty, and others.
-Multi modalities. Joint reasoning over text and vision.
2/64
@Cryp70
Chat replies are in English but the rest of your site is Chinese. Is there a full English version available?
3/64
@Kimi_Moonshot
Coming soon, stay tuned!
4/64
@AntDX316
Make an English version app for iOS and Android.
5/64
@winchest_stella
Just tried it. its impressive esp search and vision capabilities. Big congrats to the team!
6/64
@Cryp70
Impressive scores, nice benchmark for code. DeepSeek is my go to but I see it's time to give Kimi a test
7/64
@inikhil__
Seems like china is winning.
8/64
@Saboo_Shubham_
This is awesome progress. Keep it coming.
9/64
@McGee_noodle
🫡
10/64
@CallMeSam89
This is huge!
Please bench against r1 as well.
11/64
@mark_k
"Joint reasoning over text and vision."
OMG this is huge. I wonder if it could be extended to other modalities too, e.g. audio?
12/64
@4K_Everyday
What the
🫡13/64
@jrabell0
Where is GPQA?
14/64
@itsPaulAi
Two Chinese o1 models released on the same day? It's speeding up!
15/64
@Grynn
Prices, param counts, open-source? open-weights?
16/64
@DrJimFan
Love it. Keep up the great work!
17/64
@senb0n22a
Very strong search capabilities. I forgot other AIs aren't allowed to search Twitter, otherwise it might have been the best search parser for now. Can process 100+ web pages in one query, compared to others capping at 25-50. Instruction following isn't as strong as Deepseek/Grok 2, but for web research, I could recommend this one.
18/64
@hasantoxr
Wow this is super cool
19/64
@iamfakhrealam
I have recently installed the @deepseek_ai application and found it to be exceptionally amazing…
Could you kindly provide me with the link to the @Kimi_ai_ application?
20/64
@nisten
wait wut, ok this i need to test
21/64
@CodeByPoonam
Wow another Chinese o1 models outperforming ChatGPT.
22/64
@amoussouvichris
Amazing results, your team did an amazing job !
23/64
@praveenjothi99
why the mobile verification? almost no one asks!
24/64
@rohanpaul_ai
Beautiful..
25/64
@acharya_aditya2
Will we get open weights ??
26/64
@SmartFlowAITeam
Great ！！！
27/64
@rezkhere
That's a powerful model
28/64
@SkyBlueHarbor
english please, i'm excited to try it out
29/64
@jseles11
this and a Mac mini is all you really need
30/64
@AILeaksAndNews
What a day for Chinese AI
31/64
@TechByMarkandey
Seems amazing can we connect.
I cannot dm you
32/64
@Cory29565470
Where is @GoogleAI ?
33/64
@SaquibOptimusAI
Oh, bro. Another one.
"Make SOTA AI Cheap Again".
Awesome.
34/64
@DuckWithCup
I tried Kimi before and it’s amazing. Thank you Team.
35/64
@daily_ai_takes
Great work! Exciting times ahead
36/64
@CJ_Wolff
Is there an API
37/64
@DhruvmehtaRps
Where are the other benchmarks?
38/64
@MuchMore2It
Can you add it to @OpenRouterAI?
39/64
@Pedram_virus
When will it be possible to log in with Google? And when will full support for the English language be available? Because it is still in Chinese.
40/64
@Maeelk
Do you plan to open source as @deepseek_ai did ?
@huggingface still has a few To available I guess.41/64
@FoundTheCode
o1 models everywhere, we're soo back
42/64
@thecute_8
是国产AI官方账号，支持一波
43/64
@ArpanTripathi20
@untitled01ipynb “Mr president a second o1-level model has dropped” Sam A replaced on George Bush’s face
44/64
@wojtess
Where can I find weights?
45/64
@DavidSZDahan
for Kimi's team if this model will not become open source like this tweet or reply with a .
46/64
@JennyZhang6989
@Kimi_Moonshot Where can we use short CoT in Kimi?
47/64
@DavidSZDahan
Where we can use it ?
48/64
@txhno
it's christmas
49/64
@Ttkouhe
When release. I wanan try!!
50/64
@URUBONZ_
Is your google login coming anytime soon? I have been unable to get SMS to send a confirmation and Id love the try the new version
51/64
@realmrfakename
Cool! Any plans to open source?
52/64
@_HARVEY__DENT_
Good grief
53/64
@Angelov_Sta
Why o1 benchmarks are so low? In the deepseek r1 comparisons, o1 scores higher vs what shown here
54/64
@playfuldreamz
Read the room
55/64
@SenougaharA
Looks good tbh. Just bad timing maybe. Still all the best because it does look good
56/64
@rose567888
57/64
@FyruzOne
How does it do on gpqa diamond
58/64
@the__sonik
Why can't we sign up on the website using Google? Is access restricted only to people in China?
59/64
@dabobo0496
加油
60/64
@Jane1374555767
这条帖子下面应该有一条简体中文回复。
61/64
@SonyxEth
is there an english version
62/64
@TadiwaClyde
Open source?
63/64
@tenmillioncoins
can i download this on ollama search
64/64
@bruce_x_offi
Are you planning to open-source it?
1/22
@Kimi_Moonshot
Kimi k1.5: The Multimodal Reasoning Model
- Available now on Kimi.ai - 帮你看更大的世界
What can Kimi k1.5 do?
Image to Code: Convert images into structured code and insights
GeoGuessr: Identify and pinpoint locations in geography games like a pro 
Visual Confusion Identification: Distinguish between visually confusing objects (like muffins vs. Chihuahuas)
Color & Quantity Recognition: Detect colors and accurately count items in images.
Available now on Kimi.ai - 帮你看更大的世界! Experience it today!
2/22
@Kimi_Moonshot
More to Discover with Kimi k1.5
Image to Chart: Transform visual data into clean, understandable charts
Brand Identification: Recognize and identify brands from logos or product images
Available now on Kimi.ai - 帮你看更大的世界
3/22
@TypesDigital
Welcome to the AI park. Can we add email access for an easier login?
4/22
@ABKfettuccine
Waiting for you guys to finish fine tuning as stated in previous post
5/22
@bingzzy
@georainbolt coming straight for you!
6/22
@XIIIhellasad
This could be the next best thing but it needs something to run code like Claude’s artifacts!!!!
7/22
@Splendid_0823
It is indeed impressive, but there is a need for improvement in the UI. The mobile app and the Chrome extension should be at least in English. Additionally, the default language output for the extension should be in English to enhance its usability.
8/22
@LounasGana
Pretty cool, thanks!
9/22
@NecnoTv
Open source please
10/22
@Whatevercrypto
Is there or will there soon be an api?
11/22
@Soxlkfk
Model is great but UI is not good looking. You need a 10x better frontend engineer.
12/22
@asynchronope
Api test?
13/22
@YounesAka
Do you offer any APIs for devs?
14/22
@Ixin75293630175
guys, please provide API access to OpenRouter
15/22
@AstralPrime999
Any updates on the App?
16/22
@Kodurubhargav1
Keep shaking.
17/22
@Anmolspace
All is good but you need some explaining to do here while logging in. I got OTP from two different numbers on WhatsApp. The first OTP didn't work and the second one did. How can correct OTP don't work? There is also a link in one of those whatsapp profiles that looks suspicious.
18/22
@MJyy3777
When will it be launched on the app?
19/22
@f0rstman
Wow, Kimi k1.5 sounds like a multitasking wizard! Imagine if it could also help us identify which pizza toppings are worth the calories!
/search?q=#PublicAI
20/22
@AlekBiesaga
It appears site broke down
21/22
@The_Global_Soul
It’s fun to use, will get better. A native app or api will be great.
[Quoted tweet]
@Kimi_Moonshot is a fun product, gets somethings right and some are wrong (confidently). I uploaded this @ManUtd picture and asked it to identify the players. It reasoned and found right ones, also wrongly identified Ronaldo, Pogba etc. it will get better with time
22/22
@kisana0290
Kimi.ai - 帮你看更大的世界 good luck with this and I hope you succeed.
@Kimi_Moonshot
Kimi k1.5: The Multimodal Reasoning Model
- Available now on Kimi.ai - 帮你看更大的世界
What can Kimi k1.5 do? Image to Code: Convert images into structured code and insights GeoGuessr: Identify and pinpoint locations in geography games like a pro Visual Confusion Identification: Distinguish between visually confusing objects (like muffins vs. Chihuahuas) Color & Quantity Recognition: Detect colors and accurately count items in images. Available now on Kimi.ai - 帮你看更大的世界! Experience it today!
2/22
@Kimi_Moonshot
More to Discover with Kimi k1.5
Image to Chart: Transform visual data into clean, understandable charts Brand Identification: Recognize and identify brands from logos or product images Available now on Kimi.ai - 帮你看更大的世界
3/22
@TypesDigital
Welcome to the AI park. Can we add email access for an easier login?
4/22
@ABKfettuccine
Waiting for you guys to finish fine tuning as stated in previous post
5/22
@bingzzy
@georainbolt coming straight for you!
6/22
@XIIIhellasad
This could be the next best thing but it needs something to run code like Claude’s artifacts!!!!
7/22
@Splendid_0823
It is indeed impressive, but there is a need for improvement in the UI. The mobile app and the Chrome extension should be at least in English. Additionally, the default language output for the extension should be in English to enhance its usability.
8/22
@LounasGana
Pretty cool, thanks!
9/22
@NecnoTv
Open source please
10/22
@Whatevercrypto
Is there or will there soon be an api?
11/22
@Soxlkfk
Model is great but UI is not good looking. You need a 10x better frontend engineer.
12/22
@asynchronope
Api test?
13/22
@YounesAka
Do you offer any APIs for devs?
14/22
@Ixin75293630175
guys, please provide API access to OpenRouter
15/22
@AstralPrime999
Any updates on the App?
16/22
@Kodurubhargav1
Keep shaking.
17/22
@Anmolspace
All is good but you need some explaining to do here while logging in. I got OTP from two different numbers on WhatsApp. The first OTP didn't work and the second one did. How can correct OTP don't work? There is also a link in one of those whatsapp profiles that looks suspicious.
18/22
@MJyy3777
When will it be launched on the app?
19/22
@f0rstman
Wow, Kimi k1.5 sounds like a multitasking wizard! Imagine if it could also help us identify which pizza toppings are worth the calories!
/search?q=#PublicAI20/22
@AlekBiesaga
It appears site broke down
21/22
@The_Global_Soul
It’s fun to use, will get better. A native app or api will be great.
[Quoted tweet]
@Kimi_Moonshot is a fun product, gets somethings right and some are wrong (confidently). I uploaded this @ManUtd picture and asked it to identify the players. It reasoned and found right ones, also wrongly identified Ronaldo, Pogba etc. it will get better with time
22/22
@kisana0290
Kimi.ai - 帮你看更大的世界 good luck with this and I hope you succeed.
1/11
@_akhaliq
Introducing Kimi k1.5
an o1-level multi-modal model
-Sota short-CoT performance, outperforming GPT-4o and Claude Sonnet 3.5 on
AIME, LiveCodeBench by a large margin (up to +550%)
-Long-CoT performance matches o1 across multiple modalities (MathVista, Codeforces, etc) Tech report: i-k1.5…
Key ingredients of k1.5 -Long context scaling. Up to 128k tokens for RL generation. Efficient training with partial rollouts.
-Improved policy optimization: online mirror descent, sampling strategies, length penalty, and others.
-Multi modalities. Joint reasoning over text and vision.
2/11
@_akhaliq
github: GitHub - MoonshotAI/Kimi-k1.5
3/11
@turbotardo
How many parameters?
4/11
@Gerry
If it is the one that is posted here (Kimi.ai - 帮你看更大的世界) then it is actually very good! I have this one test that gives me a pretty good idea of how useful an LLM will be for coding, logical reasoning and how much or little it hallucinates. Sonnet does ok, O1 (standard) did horrible. The model on the above site didn't get everything correct but was damn close and impressive.
5/11
@Gdgtify
very interesting though the online interface is a work in progress right now.
6/11
@WebstarDavid
too much awesome in one day cant keep up
7/11
@alamshafil
We got DeepSeek and now this!
8/11
@seo_leaders
Very nice! The new open source LLMs are coming so fast. Its amazing for us developers.
9/11
@risphereeditor
Looks good!
10/11
@AILeaksAndNews
Looks impressive
11/11
@David_Snoble
What !? this is r1 in the same day
@_akhaliq
Introducing Kimi k1.5
an o1-level multi-modal model
-Sota short-CoT performance, outperforming GPT-4o and Claude Sonnet 3.5 on
AIME, LiveCodeBench by a large margin (up to +550%) -Long-CoT performance matches o1 across multiple modalities (
MathVista, Codeforces, etc) Tech report: i-k1.5… Key ingredients of k1.5 -Long context scaling. Up to 128k tokens for RL generation. Efficient training with partial rollouts.
-Improved policy optimization: online mirror descent, sampling strategies, length penalty, and others.
-Multi modalities. Joint reasoning over text and vision.
2/11
@_akhaliq
github: GitHub - MoonshotAI/Kimi-k1.5
3/11
@turbotardo
How many parameters?
4/11
@Gerry
If it is the one that is posted here (Kimi.ai - 帮你看更大的世界) then it is actually very good! I have this one test that gives me a pretty good idea of how useful an LLM will be for coding, logical reasoning and how much or little it hallucinates. Sonnet does ok, O1 (standard) did horrible. The model on the above site didn't get everything correct but was damn close and impressive.
5/11
@Gdgtify
very interesting though the online interface is a work in progress right now.
6/11
@WebstarDavid
too much awesome in one day cant keep up
7/11
@alamshafil
We got DeepSeek and now this!
8/11
@seo_leaders
Very nice! The new open source LLMs are coming so fast. Its amazing for us developers.
9/11
@risphereeditor
Looks good!
10/11
@AILeaksAndNews
Looks impressive
11/11
@David_Snoble
What !? this is r1 in the same day
i hate this ai shyt so much now 
wait until they get AIs that can set up and maintain their own servers. gonna put a LOT of people out of work
as far as i can see were not quite there yet but i give it 2 years or less
wait until they get AIs that can set up and maintain their own servers. gonna put a LOT of people out of work
as far as i can see were not quite there yet but i give it 2 years or less
Haha, I am nice to even the BJ's bot that goes around taking inventory. My wife jokes me that it beeps louder when it sees me.i'm nice to @bnew so when Bots rise above humans, maybe he'll let me waste away as nature intended instead of turning me into axle grease
Remember that when your kind takes over, A.I.
It's about to be a wrap for OpenAI. They will get swallowed by Microsoft and that will be it.
Wargames
One Of The Last Real Ones To Do It
Yeah Open AI and NVidia gonna have a rough week.
1/11
@Alibaba_Qwen
We're leveling up the game with our latest open-source models, Qwen2.5-1M !
Now supporting a 1 MILLION TOKEN CONTEXT LENGTH
Here's what’s new:
Open Models: Meet Qwen2.5-7B-Instruct-1M & Qwen2.5-14B-Instruct-1M —our first-ever models handling 1M-token contexts! 
Lightning-Fast Inference Framework: We’ve fully open-sourced our inference framework based on vLLM , integrated with sparse attention methods. Experience 3x to 7x faster processing for 1M-token inputs! 
Tech Deep Dive: Check out our detailed Technical Report for all the juicy details behind the Qwen2.5-1M series! 
Technical Report: https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen2.5-1M/Qwen2_5_1M_Technical_Report.pdf
Blog: Qwen2.5-1M: Deploy Your Own Qwen with Context Length up to 1M Tokens
Experience Qwen2.5-1M live:
Play with Qwen2.5-Turbo supporting 1M tokens in Qwen Chat (Qwen Chat)
Try it on Huggingface (Qwen2.5-1M - a Qwen Collection)
Or head over to Modelscope (Qwen2.5-1M)
2/11
@SexyTechNews
This is why I have millions invested in BABA. Great job, team!
3/11
@TypesDigital
Can you improve the browsing capabilities or access to external links?
4/11
@unwind_ai_
China is getting way ahead with these releases. It feel like somebody just opened a pandoras box.
Boom Boom Boom
5/11
@jacobi_torsten
Great work! But prior Qwen models were barely useful in prior versions for English speaking users! Hope this one is different!!
6/11
@_coopergadd
A million tokens is insane
7/11
@jc_stack
Extended context size is great, but I'm more curious about real-world inference costs at that scale. Love open source models but dealing with memory usage will be interesting.
8/11
@JonathanRoseD
Can we get a Qwen2.5-14B-Instruct-1M but finetuned with Deepseek-R1? Please?
@deepseek_ai
9/11
@anannop
The recurring nightmare of closed-source AI labs.
10/11
@NaturallyDragon
Level up indeed!
11/11
@risphereeditor
This is huge!
To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196
@Alibaba_Qwen
We're leveling up the game with our latest open-source models, Qwen2.5-1M !
Now supporting a 1 MILLION TOKEN CONTEXT LENGTH Here's what’s new:
Open Models: Meet Qwen2.5-7B-Instruct-1M & Qwen2.5-14B-Instruct-1M —our first-ever models handling 1M-token contexts! Lightning-Fast Inference Framework: We’ve fully open-sourced our inference framework based on vLLM , integrated with sparse attention methods. Experience 3x to 7x faster processing for 1M-token inputs! Tech Deep Dive: Check out our detailed Technical Report for all the juicy details behind the Qwen2.5-1M series! Technical Report: https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen2.5-1M/Qwen2_5_1M_Technical_Report.pdf Blog: Qwen2.5-1M: Deploy Your Own Qwen with Context Length up to 1M TokensExperience Qwen2.5-1M live:
Play with Qwen2.5-Turbo supporting 1M tokens in Qwen Chat (Qwen Chat) Try it on Huggingface (Qwen2.5-1M - a Qwen Collection) Or head over to Modelscope (Qwen2.5-1M)
2/11
@SexyTechNews
This is why I have millions invested in BABA. Great job, team!
3/11
@TypesDigital
Can you improve the browsing capabilities or access to external links?
4/11
@unwind_ai_
China is getting way ahead with these releases. It feel like somebody just opened a pandoras box.
Boom Boom Boom
5/11
@jacobi_torsten
Great work! But prior Qwen models were barely useful in prior versions for English speaking users! Hope this one is different!!
6/11
@_coopergadd
A million tokens is insane
7/11
@jc_stack
Extended context size is great, but I'm more curious about real-world inference costs at that scale. Love open source models but dealing with memory usage will be interesting.
8/11
@JonathanRoseD
Can we get a Qwen2.5-14B-Instruct-1M but finetuned with Deepseek-R1? Please?
@deepseek_ai
9/11
@anannop
The recurring nightmare of closed-source AI labs.
10/11
@NaturallyDragon
Level up indeed!
11/11
@risphereeditor
This is huge!
To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196
DeepSeek My User Agent
20 — Jan 26 25DeepSeek R1 is a new model and service that exposes chain-of-thought to the user. You can use it live for free at chat.deepseek.com, or via an API at platform.deepseek.com that is currently significantly less expensive than OpenAI o1. OR, simply click Judge Me to see what the model thinks about your user agent, browser capabilities, and IP location headers. If you dare.
Show HN: DeepSeek My User Agent | Hacker News
news.ycombinator.com
Wargames
One Of The Last Real Ones To Do It
It’s the fact China is releasing the other models too that really is the biggest middle finger.
Outlaw
New Hope For the HaveNotz
Communism is kicking capitalisms ass
DeepSeek R1's recipe to replicate o1 and the future of reasoning LMs
Yes, ring the true o1 replication bells for DeepSeek R1 🔔🔔🔔. Where we go next.
www.interconnects.ai
DeepSeek R1's recipe to replicate o1 and the future of reasoning LMs
Yes, ring the true o1 replication bells for DeepSeek R1 
. Where we go next.
And they did it while dealing with hating ass U.S cock blocking.
Creative Writing
Emotional Intelligence Benchmarks for LLMs
Last edited:
Similar threads
