NanoGPT
@nanogpt.bsky.social
Access every AI model privately - no subscription, pay only for what you use.
The Nous 4 models have just been added: Large (405b) and Medium (70b).
These models have a hybrid thinking mode, with the model being able to reason internally. It's also a model that has a far lower refusal rate than most models have - lower even than Grok 4!
These models have a hybrid thinking mode, with the model being able to reason internally. It's also a model that has a far lower refusal rate than most models have - lower even than Grok 4!
August 26, 2025 at 7:55 PM
The Nous 4 models have just been added: Large (405b) and Medium (70b).
These models have a hybrid thinking mode, with the model being able to reason internally. It's also a model that has a far lower refusal rate than most models have - lower even than Grok 4!
These models have a hybrid thinking mode, with the model being able to reason internally. It's also a model that has a far lower refusal rate than most models have - lower even than Grok 4!
Folders are now live!
Go to the /conversations tab to create folders, then use them from the /conversations page or from the "most recent conversations" dropdown.
You of course can also still show "all" conversations, and search across all conversations.
Go to the /conversations tab to create folders, then use them from the /conversations page or from the "most recent conversations" dropdown.
You of course can also still show "all" conversations, and search across all conversations.
August 26, 2025 at 3:55 PM
Folders are now live!
Go to the /conversations tab to create folders, then use them from the /conversations page or from the "most recent conversations" dropdown.
You of course can also still show "all" conversations, and search across all conversations.
Go to the /conversations tab to create folders, then use them from the /conversations page or from the "most recent conversations" dropdown.
You of course can also still show "all" conversations, and search across all conversations.
Importing ChatGPT and Claude conversations is now live.
If you've been postponing moving to NanoGPT because of your extensive ChatGPT/Claude history - export it, import it, and increase your privacy and flexibility.
2-step instructions below!
If you've been postponing moving to NanoGPT because of your extensive ChatGPT/Claude history - export it, import it, and increase your privacy and flexibility.
2-step instructions below!
NanoGPT | Every AI model | Privacy-first | No subscription | Text, Image, Video
Access all the newest AI models including ChatGPT, Claude, Gemini, Deepseek, and all image and video models. Starting at $0.01, pay only for what you use. Private, no subscription needed.
nano-gpt.com
August 26, 2025 at 11:18 AM
Importing ChatGPT and Claude conversations is now live.
If you've been postponing moving to NanoGPT because of your extensive ChatGPT/Claude history - export it, import it, and increase your privacy and flexibility.
2-step instructions below!
If you've been postponing moving to NanoGPT because of your extensive ChatGPT/Claude history - export it, import it, and increase your privacy and flexibility.
2-step instructions below!
You can now queue messages in our chat!
Helpful if you're thinking of something while the model is outputting and want to queue it for sending immediately, or for example if you're fond of Barry the Bear and want the story to continue, and continue...
Queued messages can also be deleted.
Helpful if you're thinking of something while the model is outputting and want to queue it for sending immediately, or for example if you're fond of Barry the Bear and want the story to continue, and continue...
Queued messages can also be deleted.
August 24, 2025 at 6:57 AM
You can now queue messages in our chat!
Helpful if you're thinking of something while the model is outputting and want to queue it for sending immediately, or for example if you're fond of Barry the Bear and want the story to continue, and continue...
Queued messages can also be deleted.
Helpful if you're thinking of something while the model is outputting and want to queue it for sending immediately, or for example if you're fond of Barry the Bear and want the story to continue, and continue...
Queued messages can also be deleted.
Want to look up old conversations or image generations? It's now possible to sort by purely text conversations on the /conversations page, or check out just your favorites/pinned conversations, or of course go to the gallery to see all the images and videos you've generated.
August 23, 2025 at 6:20 PM
Want to look up old conversations or image generations? It's now possible to sort by purely text conversations on the /conversations page, or check out just your favorites/pinned conversations, or of course go to the gallery to see all the images and videos you've generated.
Qwen Image now supports image-to-image editing!
Upload an image and use natural language to make precise edits - from semantic changes (style, layout, composition) to fine-grained adjustments (add/remove objects), all while preserving structure.
Upload an image and use natural language to make precise edits - from semantic changes (style, layout, composition) to fine-grained adjustments (add/remove objects), all while preserving structure.
August 23, 2025 at 4:47 PM
Qwen Image now supports image-to-image editing!
Upload an image and use natural language to make precise edits - from semantic changes (style, layout, composition) to fine-grained adjustments (add/remove objects), all while preserving structure.
Upload an image and use natural language to make precise edits - from semantic changes (style, layout, composition) to fine-grained adjustments (add/remove objects), all while preserving structure.
Prices for both variants of the open-source GPT OSS model (OpenAI's open-source model) have been decreased quite a bit!
August 23, 2025 at 3:40 PM
Prices for both variants of the open-source GPT OSS model (OpenAI's open-source model) have been decreased quite a bit!
A new and very cheap model for realistic images: CyberRealistic XL is now live!
August 23, 2025 at 12:26 PM
A new and very cheap model for realistic images: CyberRealistic XL is now live!
Our TTS (Text to Speech) offerings have been expanded! In addition to the GPT voice models, Elevenlabs, and Whisper, we now also have the Gemini 2.5 Pro and Flash text to speech models live!
August 23, 2025 at 11:21 AM
Our TTS (Text to Speech) offerings have been expanded! In addition to the GPT voice models, Elevenlabs, and Whisper, we now also have the Gemini 2.5 Pro and Flash text to speech models live!
The first open-source providers for Deepseek V3.1 are coming online, so we are no longer routing through the original provider for it but only through no-log open source providers from this point on.
Better privacy, and as an added benefit we've also lowered the price!
Better privacy, and as an added benefit we've also lowered the price!
August 21, 2025 at 11:27 AM
The first open-source providers for Deepseek V3.1 are coming online, so we are no longer routing through the original provider for it but only through no-log open source providers from this point on.
Better privacy, and as an added benefit we've also lowered the price!
Better privacy, and as an added benefit we've also lowered the price!
We've been working on improving the formatting in general, but also with a focus on how math gets output.
If you're into long equations and want them to be displayed well you should now be a lot happier with our service!
If you're into long equations and want them to be displayed well you should now be a lot happier with our service!
August 21, 2025 at 11:26 AM
We've been working on improving the formatting in general, but also with a focus on how math gets output.
If you're into long equations and want them to be displayed well you should now be a lot happier with our service!
If you're into long equations and want them to be displayed well you should now be a lot happier with our service!
Memory has been further improved - if you now try a long pre-existing conversation (so one you've never used Memory in before, but is already say 200k tokens long), they'll index it in the background rather and tell you in how long (a few minutes, usually) the conversation is indexed and ready.
August 21, 2025 at 11:26 AM
Memory has been further improved - if you now try a long pre-existing conversation (so one you've never used Memory in before, but is already say 200k tokens long), they'll index it in the background rather and tell you in how long (a few minutes, usually) the conversation is indexed and ready.
We're putting some improvements into our chat page.
It should allow for easier sorting of models, showing which support image input, which ones are reasoning models, and generally make it feel better to navigate the now 400+ models.
It should allow for easier sorting of models, showing which support image input, which ones are reasoning models, and generally make it feel better to navigate the now 400+ models.
August 20, 2025 at 12:59 PM
We're putting some improvements into our chat page.
It should allow for easier sorting of models, showing which support image input, which ones are reasoning models, and generally make it feel better to navigate the now 400+ models.
It should allow for easier sorting of models, showing which support image input, which ones are reasoning models, and generally make it feel better to navigate the now 400+ models.
Context memory now live!
TL:DR: we've added context memory which gives infinite memory/context size to any model and improves recall, speed, and performance.
This is an amazing feature for programming and agentic purposes, and can be used with any model.
www.youtube.com/watch?v=ecTj...
TL:DR: we've added context memory which gives infinite memory/context size to any model and improves recall, speed, and performance.
This is an amazing feature for programming and agentic purposes, and can be used with any model.
www.youtube.com/watch?v=ecTj...
Context Memory
www.youtube.com
August 13, 2025 at 2:01 PM
Context memory now live!
TL:DR: we've added context memory which gives infinite memory/context size to any model and improves recall, speed, and performance.
This is an amazing feature for programming and agentic purposes, and can be used with any model.
www.youtube.com/watch?v=ecTj...
TL:DR: we've added context memory which gives infinite memory/context size to any model and improves recall, speed, and performance.
This is an amazing feature for programming and agentic purposes, and can be used with any model.
www.youtube.com/watch?v=ecTj...
We're temporarily discounting the Qwen Image model! Now just $0.01 per image!
August 11, 2025 at 11:16 AM
We're temporarily discounting the Qwen Image model! Now just $0.01 per image!
It's now also possible to do embeddings on our API!
If you don't know what this is - see next tweet.
For those that do and want to do embedding privately, check out the documentation here docs.nano-gpt.com/api-referenc...
We support 15 embedding models right now.
If you don't know what this is - see next tweet.
For those that do and want to do embedding privately, check out the documentation here docs.nano-gpt.com/api-referenc...
We support 15 embedding models right now.
Embeddings - NanoGPT API Documentation
Create embeddings for text using OpenAI-compatible and alternative embedding models
docs.nano-gpt.com
August 11, 2025 at 6:51 AM
It's now also possible to do embeddings on our API!
If you don't know what this is - see next tweet.
For those that do and want to do embedding privately, check out the documentation here docs.nano-gpt.com/api-referenc...
We support 15 embedding models right now.
If you don't know what this is - see next tweet.
For those that do and want to do embedding privately, check out the documentation here docs.nano-gpt.com/api-referenc...
We support 15 embedding models right now.
Qwen Image is live!
This 20B image model by Alibaba's Qwen is great at complex text rendering - supporting multi-line layouts, paragraph-level semantics, and bilingual content (English & Chinese).
This 20B image model by Alibaba's Qwen is great at complex text rendering - supporting multi-line layouts, paragraph-level semantics, and bilingual content (English & Chinese).
August 7, 2025 at 8:46 AM
Qwen Image is live!
This 20B image model by Alibaba's Qwen is great at complex text rendering - supporting multi-line layouts, paragraph-level semantics, and bilingual content (English & Chinese).
This 20B image model by Alibaba's Qwen is great at complex text rendering - supporting multi-line layouts, paragraph-level semantics, and bilingual content (English & Chinese).
Reposted by NanoGPT
NanoGPT - Monero takes the crown, another month
Our July Payment statistics are out!
nano-gpt.com/blog/july-pa...
With #1 Monero, #2 Nano and #3 Bitcoin.
Read more in the blog post!
nano-gpt.com/blog/july-pa...
With #1 Monero, #2 Nano and #3 Bitcoin.
Read more in the blog post!
July Payment Statistics
Analysis of cryptocurrency payment distribution for July 2025
nano-gpt.com
August 1, 2025 at 5:54 PM
NanoGPT - Monero takes the crown, another month
NanoGPT now also supports Study Mode - ChatGPT with a custom instruction to help with learning.
AI models are quickly becoming some of the most widely used tools in the world. But it does sometimes feel like we're just offloading our thinking, rather than using it to learn.
AI models are quickly becoming some of the most widely used tools in the world. But it does sometimes feel like we're just offloading our thinking, rather than using it to learn.
August 2, 2025 at 11:21 AM
NanoGPT now also supports Study Mode - ChatGPT with a custom instruction to help with learning.
AI models are quickly becoming some of the most widely used tools in the world. But it does sometimes feel like we're just offloading our thinking, rather than using it to learn.
AI models are quickly becoming some of the most widely used tools in the world. But it does sometimes feel like we're just offloading our thinking, rather than using it to learn.
ArliAI RPR Ultra 235B has been added.
Interesting one for our roleplaying users - ArliAI is a provider that specifically hosts many roleplaying finetunes, and this is their first custom-tuned larger roleplaying model at 235B parameters.
Interesting one for our roleplaying users - ArliAI is a provider that specifically hosts many roleplaying finetunes, and this is their first custom-tuned larger roleplaying model at 235B parameters.
August 2, 2025 at 7:10 AM
ArliAI RPR Ultra 235B has been added.
Interesting one for our roleplaying users - ArliAI is a provider that specifically hosts many roleplaying finetunes, and this is their first custom-tuned larger roleplaying model at 235B parameters.
Interesting one for our roleplaying users - ArliAI is a provider that specifically hosts many roleplaying finetunes, and this is their first custom-tuned larger roleplaying model at 235B parameters.
Our July Payment statistics are out!
nano-gpt.com/blog/july-pa...
With #1 Monero, #2 Nano and #3 Bitcoin.
Read more in the blog post!
nano-gpt.com/blog/july-pa...
With #1 Monero, #2 Nano and #3 Bitcoin.
Read more in the blog post!
July Payment Statistics
Analysis of cryptocurrency payment distribution for July 2025
nano-gpt.com
August 1, 2025 at 3:26 PM
Our July Payment statistics are out!
nano-gpt.com/blog/july-pa...
With #1 Monero, #2 Nano and #3 Bitcoin.
Read more in the blog post!
nano-gpt.com/blog/july-pa...
With #1 Monero, #2 Nano and #3 Bitcoin.
Read more in the blog post!
Step-3 now available on NanoGPT!
This multimodal reasoning model from StepFun features a Mixture-of-Experts architecture with 321B total parameters and 38B active.
See benchmarks in the quoted tweet!
This multimodal reasoning model from StepFun features a Mixture-of-Experts architecture with 321B total parameters and 38B active.
See benchmarks in the quoted tweet!
July 31, 2025 at 3:22 PM
Step-3 now available on NanoGPT!
This multimodal reasoning model from StepFun features a Mixture-of-Experts architecture with 321B total parameters and 38B active.
See benchmarks in the quoted tweet!
This multimodal reasoning model from StepFun features a Mixture-of-Experts architecture with 321B total parameters and 38B active.
See benchmarks in the quoted tweet!
Two new Wan 2.2 video models are now available! Wan 2.2 Turbo (nano-gpt.com/media?mode=v...) and Wan 2.2 5b are faster and more affordable versions of the Wan 2.2 14b model.
Both support text-to-video and image-to-video generation. Wan 2.2 Turbo starts at just $0.05 per video for 480p!
Both support text-to-video and image-to-video generation. Wan 2.2 Turbo starts at just $0.05 per video for 480p!
NanoGPT | Every AI model | Privacy-first | No subscription | Text, Image, Video
Access all the newest AI models including ChatGPT, Claude, Gemini, Deepseek, and all image and video models. Starting at $0.01, pay only for what you use. Private, no subscription needed.
nano-gpt.com
July 31, 2025 at 1:00 PM
Two new Wan 2.2 video models are now available! Wan 2.2 Turbo (nano-gpt.com/media?mode=v...) and Wan 2.2 5b are faster and more affordable versions of the Wan 2.2 14b model.
Both support text-to-video and image-to-video generation. Wan 2.2 Turbo starts at just $0.05 per video for 480p!
Both support text-to-video and image-to-video generation. Wan 2.2 Turbo starts at just $0.05 per video for 480p!
New model: Qwen3 30B A3B 2507 is now available!
nano-gpt.com/conversation...
This 30.5B-parameter mixture-of-experts language model from Qwen features 3.3B active parameters per inference, offering excellent efficiency. Significant improvements compared to its predecessor!
nano-gpt.com/conversation...
This 30.5B-parameter mixture-of-experts language model from Qwen features 3.3B active parameters per inference, offering excellent efficiency. Significant improvements compared to its predecessor!
NanoGPT | Every AI model | Privacy-first | No subscription | Text, Image, Video
Access all the newest AI models including ChatGPT, Claude, Gemini, Deepseek, and all image and video models. Starting at $0.01, pay only for what you use. Private, no subscription needed.
nano-gpt.com
July 30, 2025 at 8:11 PM
New model: Qwen3 30B A3B 2507 is now available!
nano-gpt.com/conversation...
This 30.5B-parameter mixture-of-experts language model from Qwen features 3.3B active parameters per inference, offering excellent efficiency. Significant improvements compared to its predecessor!
nano-gpt.com/conversation...
This 30.5B-parameter mixture-of-experts language model from Qwen features 3.3B active parameters per inference, offering excellent efficiency. Significant improvements compared to its predecessor!
Just 3 more days of Veo 3 Fast being discounted! The no audio version of Veo 3 Fast is just $1.20 per video until the end of July, then the price will increase to the regular $2 per video.
July 29, 2025 at 7:47 AM
Just 3 more days of Veo 3 Fast being discounted! The no audio version of Veo 3 Fast is just $1.20 per video until the end of July, then the price will increase to the regular $2 per video.