A request is any message you send to the chatbot. Each month you have a fixed package of requests available. Within it, you can communicate with AI. For example, simple AI models spend one request from your package per your message. More advanced neural networks may use, for example, three requests for one message. This is because they: Are more powerful and smarter Require more computational resources to generate a quality response Have higher cost price and final price

What to do if requests run out?

There are two options: If you regularly hit the limit and understand it won't be enough going forward, we recommend switching to a higher plan If you need more requests only occasionally, you can purchase them separately

What neural networks do you have?

We add almost all models that can be useful in everyday life for most users. We monitor new releases and add them promptly — usually within one or two days. We also keep old versions of neural networks for a while in case the new one proves unstable. The full and up-to-date list of available neural networks can be viewed here

How to cancel the subscription?

You can cancel your subscription at any time: In the bot: use the /unsubscribe command and follow the instructions In the web version: click on your profile icon in the top right corner, go to settings, and find the “Unsubscribe” button But it's better not to cancel anything, of course, because soon robots will take over the world, and we don't see much point in resisting 😉

Do you have technical support?

Of course, it can be reached either by email at support@chataibot.ru or @ChatAISupport

Can multiple people be added to the subscription?

Yes, this option is available in the “Business” plan. Up to 30 users can be added to it. Despite the name, any individual can subscribe to this plan. It's perfect for a family or small team, for example.

Who has access to my chats, photos, and videos?

No one except you, us, and the neural networks you use. We don't grant access to anyone else; your privacy is under your control.

Price - Chat AI

Choose a pricing plan for your tasks

Connect premium functionality and expand your capabilities. Access is granted on the on the site, in the in the Telegram bot, in the app and in the in the browser extension.

Free

For the simplest requests

0 €
Start for free
- Monthly requests
  
  500
- Message length
  
  2 500 symbols
- Work on a convenient platform
  
  Website
  
  Browser extension
  
  Telegram bot
  
  Application
- Limited access to base models
- Internet search
  
  Available models for text:
  
  GPT‑4o Search
  Perplexity
  
  Gemini 3.5 Flash Search
- Image and logo generation + text overlay
  
  Available models for image:
  
  GPT Image 1
  GPT Image 1.5
  Google Nano Banana
  Ideogram 3 Turbo
  Ideogram 3
  Ideogram 4
  Midjourney 6.1
  Midjourney 7
  Flux 1.1 Pro
  Flux Schnell
  Flux Ultra
  Flux Kontext Max
  Grok Imagine
  Grok Imagine Quality
  Recraft v3
  Seedream 4.0
  Seedream 5.0 Lite
  Qwen Lora
  
  GPT Image 2
  
  Google Nano Banana 2
  
  Ideogram 4 Turbo
  
  Midjourney 8.1
- Solving problems using photographs, as well as analyzing images and documents
- Convenient organization of chats for any projects
The most popular
Premium 💎

Optimal for studying and working

750 € /month

750 €

Start
- Monthly requests
  
  When paying for 3, 6 or 12 months, requests are combined and can be used at any time
  
  2 500
- Message length
  
  100 000 symbols
- Everything from the free plan, plus:
- Full access to GPT-5.5 and other everyday models
  
  Available models for text:
  
  GPT-4.1
  GPT-5.4
  GPT-5.4 mini
  GPT-5.5 High
  Claude Haiku 4.5
  Claude Sonnet 4.6
  Gemini 2.5 Pro
  Gemini 3.0 Flash
  DeepSeek-R1
  DeepSeek-V3.2
  Grok 4.3 xAI
  Qwen3
  Qwen3.5
  Qwen3.5 Plus
- In-depth expert answers
  
  Available models for text:
  
  GPT-5.2 High
  Gemini 3.1 Pro
  Claude Opus 4.7
  Claude Sonnet High 4.6
- Access to advanced models for image generation
  
  GPT Image 2 High
  
  Google Nano Banana Pro
  
  MICROSOFT-mai-image-2.5
- Creating a video from text
  
  Available models for video:
  
  KLING Standard
  KLING Pro
  Seedance Lite
  Grok Imagine Video
  
  KLING O1
  
  Seedance 2.0 Fast
  
  Gemini Omni
- Unlimited ♾️
  
  GPT-4.1 nano
- Improved search with source verification
  
  Perplexity Pro
- Voice assistant and document analysis
- Create your own AI assistants for routine tasks
Pro 🚀

For professional use

2 000 € /month

2 000 €

Start
- Monthly requests
  
  When paying for 3, 6 or 12 months, requests are combined and can be used at any time
  
  7 000
- Message length
  
  2 000 000 symbols
- All Premium features, plus:
- Flagship models
  
  o3-pro
  
  GPT-5.4 Pro
  
  Claude Opus 4.1
- Professional videos for social media
  
  Available models for video:
  
  KLING 2.0 Master
  Seedance Pro
  
  Sora 2
  
  Kling 3.0 Standard
  
  Seedance 2.0
- More requests - more freedom
Business 💼

The best option for teams and video makers

5 000 € /month

5 000 €

Start

Choose a convenient way to contact

Write an email Write in Telegram
- Monthly requests
  
  When paying for 3, 6 or 12 months, requests are combined and can be used at any time
  
  20 000
- Message length
  
  2 000 000 symbols
- All Pro features, plus:
- Personal manager
- Access to all models we offer
- One subscription for the whole team
- Employee expense management
- Create your own AI assistants based on your company
- High-quality videos with sound support
  
  Available models for video:
  
  Google VEO 3.1
  Google VEO 3.1 Fast
  Sora 2 Pro
  KLING 3.0 Pro

Full feature matrix

Select plan	Free 0 € / /month Start for free	Premium 💎 750 € / /month 750 ₽ Start	Pro 🚀 2 000 € / /month 2 000 ₽ Start	Business 💼 5 000 € / /month 5 000 ₽ Start
Main features
💬 Monthly requests	450	2 500	7 000	20 000
📩 Maximum message length in characters	2 500	100 000	2 000 000	2 000 000
🧑‍💼 Voice assistant
📷 Photo recognition
📝 Problem solving
🤖 Create an assistant in Telegram
Platforms
Website
Telegram bot
Browser extension
VK bot
Android
iOS
Text models
GPT-5.4 nano GPT-5.4 nano — the smallest and cheapest version of 5.4. For simple tasks, where speed and cost are the most important Intelligence
GPT-4.1 nano A lightweight model for quick tasks, suitable for simple questions and short texts Intelligence
GPT-5.4 mini GPT-5.4 mini is the most powerful and functional among the compact models in the lineup Intelligence
Claude Haiku 4.5 An economical model that combines speed, scalability, and results Intelligence
GPT-4.1 A fast model that takes into account the dialogue context of up to 2 million characters Intelligence
DeepSeek-R1 One of the best models from China, communicates in simple language, capable of analyzing user requests Intelligence
Claude Sonnet 4.5 A smart model from Anthropic. It creates more creative and lively texts Intelligence
Claude Sonnet 4.6 The updated Sonnet model has higher efficiency and greater intelligence Intelligence
Qwen3 The model is tailored for step-by-step logic, mathematics, and is capable of reasoning using extensive context Intelligence
DeepSeek-V3.2 A large language model with high computational efficiency, reasoning abilities, and support for long context Intelligence
Gemini 3.0 Flash Combines the logical thinking capabilities of Gemini 3 Pro with the advantages of the Flash line in response speed Intelligence
GPT-5.2 An improved conversational version with a more accurate understanding of context. It solves everyday tasks even faster and handles complex ones more reliably and precisely than GPT-5.1 Intelligence
GPT-5.1 Conversational version with adaptive thinking. Solves simple tasks instantly, analyzes complex ones in detail Intelligence
Gemini 2.5 Pro The smartest model from Google Intelligence
Gemini 3.5 Flash An optimized Google model that combines advanced intelligence, high speed, and cost-effectiveness for a wide range of practical tasks Intelligence
GPT-5.4 The new advanced model for complex professional tasks ensures high accuracy and helps solve problems quickly Intelligence
Qwen3.5 The Qwen model features enhanced performance and efficiency, comparable to leading models Intelligence
Grok 4.3 xAI A powerful model from xAI with 1M context, reasoning support, and high scores in benchmarks Intelligence
GPT-5.5 The latest flagship model for the most complex professional tasks guarantees the highest precision and quality of results Intelligence
Claude Sonnet High 4.5 Improved version of Claude Sonnet 4.5. The best in logical analysis, solving mathematical problems, and processing complex data Intelligence
Claude Sonnet High 4.6 The advanced version of Sonnet 4.6. It works better with context analysis, engages in more reasoning when solving problems, and processing complex data Intelligence
Qwen3 Max The largest Qwen model with enhanced accuracy in responses excels in mathematics, programming, and following instructions, supporting over a hundred languages Intelligence
GPT-5.2 High Advanced analytical version for the most complex queries. It analyzes data longer and deeper, constructs long chains of reasoning, and selects the most substantiated solutions Intelligence
GPT-5.1 High An analytical version with deep reflections. Thinks longer before answering to find the best solution Intelligence
Claude Fable 5 The cutting-edge model performs exceptionally in all metrics, demonstrating outstanding performance in software development Intelligence
Claude Opus 4.8 The most powerful model from Anthropic, with improved reasoning abilities, is ideal for deep analysis, development, and solving the most complex problems Intelligence
GPT-5.5 High GPT-5.5 version with an extended reasoning mode for complex tasks that require maximum accuracy and thorough elaboration Intelligence
Claude Opus 4.7 A new powerful model from Anthropic, which surpasses its predecessors, is suitable for professional use and tackling the most ambitious tasks Intelligence
GPT-5.4 High Version GPT-5.4 with enhanced analysis for complex tasks where accuracy, logic, and careful elaboration of the response are important Intelligence
Gemini 3.1 Pro The advanced logical thinking model from Google is effective for performing tasks in structured areas such as finance and programming Intelligence
Qwen3.5 Plus The extended version of Qwen 3.5, providing a deeper understanding of context and improved results in complex tasks Intelligence
Claude Opus 4.6 The flagship model Claude of the new generation with enhanced intelligence for programming, stable operation in large codebases, and improved debugging and code review Intelligence
Claude Opus 4.5 The latest flagship model, Claude. Unmatched capabilities in solving the most complex problems with improved speed and accuracy Intelligence
Claude Opus 4.1 A powerful model from Anthropic of the previous generation. High accuracy in complex tasks and deep analysis Intelligence
GPT-5.4 Pro The extended version of GPT-5.4 uses more computational resources to analyze requests more deeply Intelligence
o3-pro The refined version of o3, a leader in the market for solving the most complex problems Intelligence
Internet access
Perplexity Before responding, look for information on the internet 🌐, suitable for obtaining current data Intelligence
GPT-4o Search Model with built-in internet search 🌐 Intelligence
Gemini 3.0 Flash Search Gemini 3 Flash version with Google search support for retrieving up-to-date information Intelligence
Gemini 3.5 Flash Search The combination of speed, efficiency, and real-time search is Google's advanced intelligence with access to current information from the web Intelligence
Grok 4.3 Search xAI Grok by xAI with internet search: fresh data, fast analysis, and precise answers 🌐 Intelligence
Perplexity Pro It conducts a deeper search than the usual Perplexity Intelligence
Image generation
Flux Schnell A lightweight model, suitable for simple requests and quick responses Quality
Ideogram Turbo 4 A cheaper and faster version of Ideogram 4 Quality
Ideogram 3 Turbo Simplified version of Ideogram. Works faster with a loss of quality Quality
Ideogram 4 Design-oriented model. Quickly create high-quality visual images based on text Quality
Seedream 5.0 Lite Creating images with improved prompt understanding: more accurate, more logical, more stable — for complex creative tasks Quality
Qwen Lora a model for generating images with precise text representation, realistic people, and natural textures Quality
Seedream 4.0 Creating images at resolutions up to 4K with maximum detail and extensive creative possibilities Quality
Google Nano Banana The model from Google based on Gemini 2.5 Flash Quality
Ideogram 3 A high-quality model for creating images with text, it excels at logos Quality
GPT Image 1 Optimal balance of speed and image quality Quality
Recraft v3 Neural network for generating and editing images with high detail and support for different styles Quality
Flux Ultra Creates detailed and high-quality images, the model is based on the refined Pro version Quality
Flux 1.1 Pro The professional model from Flux is capable of competing with Midjourney in terms of quality Quality
Midjourney 6.1 The most popular model for generating images. Creates 4 images per request Quality
Midjourney 8.1 More accurate image generation, where the result aligns better with the original idea and looks more cohesive and cleaner Quality
GPT Image 2 Advanced model for fast and high-quality image generation and editing Quality
Google Nano Banana 2 Google's next-generation model: higher quality than Nano Banana, at a lower cost than Pro Quality
GPT Image 1.5 Changes only what you request without distorting details. Enhanced text rendering. Support for the Russian language Quality
Flux Kontext Max A premium model that creates thoughtful, coherent images based on the most complex texts Quality
Midjourney 7 The image quality is significantly higher, and objects of all types have much better coherence in all details Quality
Grok Imagine The image generation model from xAI is distinguished by its ability to draw well-known people without censorship Quality
Grok Imagine Quality Improved version of Grok Imagine for more detailed 2K images. Supports editing and merging up to 3 photos Quality
Microsoft MAI-Image-2.5 a model for generating images with precise text representation, realistic people, and natural textures Quality
GPT Image 2 High Higher quality generations from the GPT Image 2 model Quality
GPT Image 1.5 High Increased detail, with resolution and texture work for professional use Quality
Google Nano Banana Pro An improved model from Google based on Gemini 3 Pro, providing high-quality image generation with a resolution of 2K Quality
GPT Image 1 High It takes into account high-level details and displays the Russian text Quality
Video generation
KLING 1.6 Standard Standard version, suitable for simple frames or uncomplicated animation Quality
KLING O1 The model is capable of animating the transition between the starting and final frames based on the text, with precise movement and physics Quality
KLING 2.5 Turbo Pro Excellent generation with good motion smoothness, cinematic visual effects, and accuracy Quality
Seedance Lite Model from the creators of TikTok: instant video generation from text, suitable for simple scenes Quality
Grok Imagine Video The xAI model for creating short dynamic videos from text and images. Well-suited for photo animation and content for social media Quality
Seedance 2.0 Fast A slightly simplified version of Seedance 2.0, faster and cheaper Quality
KLING 2.6 Pro A model with cinematic graphics, smooth object movement, and built-in sound generation Quality
Gemini Omni Generation of cinematic videos in 4K resolution with realistic sound accompaniment Quality
KLING 3.0 Standard A simplified version of Kling 3.0, slightly inferior in quality but having the same capabilities Quality
Sora 2 The latest model from OpenAI creates realistic and physically accurate videos with synchronized sound and dialogues Quality
KLING 2.0 Master Creates generations with exceptional quality and realism Quality
Seedance 2.0 The most advanced model from ByteDance creates cinematic videos with synchronized sound Quality
Seedance Pro An advanced model with enhanced detail and smooth animation, perfect for social media Quality
KLING 3.0 Pro Premium image transformation into video with cinematic graphics, smooth animation, and built-in audio generation Quality
Veo 3.1 Fast A faster version of the Veo 3.1 model, which also supports sound generation Quality
Sora 2 Pro The improved version of Sora 2 generates in Full HD quality Quality
Veo 3.1 The improved version of Veo 3 creates videos composed of multiple scenes in higher quality and with context-dependent sound Quality

Full feature matrix

Main features

💬 Monthly requests

2 500

📩 Maximum message length in characters

100 000

🧑‍💼 Voice assistant

📷 Photo recognition

📝 Problem solving

🤖 Create an assistant in Telegram

Platforms

Website

Telegram bot

Browser extension

VK bot

Android

iOS

Text models

GPT-5.4 nano

GPT-4.1 nano

GPT-5.4 mini

Claude Haiku 4.5

GPT-4.1

DeepSeek-R1

Claude Sonnet 4.5

Claude Sonnet 4.6

Qwen3

DeepSeek-V3.2

Gemini 3.0 Flash

GPT-5.2

GPT-5.1

Gemini 2.5 Pro

Gemini 3.5 Flash

GPT-5.4

Qwen3.5

Grok 4.3 xAI

GPT-5.5

Claude Sonnet High 4.5

Claude Sonnet High 4.6

Qwen3 Max

GPT-5.2 High

GPT-5.1 High

Claude Fable 5

Claude Opus 4.8

GPT-5.5 High

Claude Opus 4.7

GPT-5.4 High

Gemini 3.1 Pro

Qwen3.5 Plus

Claude Opus 4.6

Claude Opus 4.5

Claude Opus 4.1

GPT-5.4 Pro

o3-pro

Internet access

Perplexity

GPT-4o Search

Gemini 3.0 Flash Search

Gemini 3.5 Flash Search

Grok 4.3 Search xAI

Perplexity Pro

Image generation

Flux Schnell

Ideogram Turbo 4

Ideogram 3 Turbo

Ideogram 4

Seedream 5.0 Lite

Qwen Lora

Seedream 4.0

Google Nano Banana

Ideogram 3

GPT Image 1

Recraft v3

Flux Ultra

Flux 1.1 Pro

Midjourney 6.1

Midjourney 8.1

GPT Image 2

Google Nano Banana 2

GPT Image 1.5

Flux Kontext Max

Midjourney 7

Grok Imagine

Grok Imagine Quality

Microsoft MAI-Image-2.5

GPT Image 2 High

GPT Image 1.5 High

Google Nano Banana Pro

GPT Image 1 High

Video generation

KLING 1.6 Standard

KLING O1

KLING 2.5 Turbo Pro

Seedance Lite

Grok Imagine Video

Seedance 2.0 Fast

KLING 2.6 Pro

Gemini Omni

KLING 3.0 Standard

Sora 2

KLING 2.0 Master

Seedance 2.0

Seedance Pro

KLING 3.0 Pro

Veo 3.1 Fast

Sora 2 Pro

Veo 3.1

Show details

Request cost by model

If the dialogue exceeds 6,000 characters — the deduction coefficient increases by ×1 for every 6,000 characters (6,001–12,000 → ×2, 12,001–18,000 → ×3, etc.).

Text models

OpenAI

GPT-5.4 nano from 1

GPT-4.1 nano from 1

GPT-5.4 mini from 3

GPT-4.1 from 3

GPT-5.2 from 4

GPT-5.1 from 3

GPT-5.4 from 6

GPT-5.5 from 12

GPT-5.2 High from 22

GPT-5.1 High from 15

GPT-5.5 High from 60

GPT-5.4 High from 30

GPT-5.4 Pro from 70

o3-pro from 25

GPT-4o Search from 15

Anthropic

Claude Haiku 4.5 from 3

Claude Sonnet 4.5 from 7

Claude Sonnet 4.6 from 7

Claude Sonnet High 4.5 from 10

Claude Sonnet High 4.6 from 10

Claude Fable 5 from 25

Claude Opus 4.8 from 12

Claude Opus 4.7 from 12

Claude Opus 4.6 from 12

Claude Opus 4.5 from 12

Claude Opus 4.1 from 20

DeepSeek

DeepSeek-R1 from 1

DeepSeek-V3.2 from 2

Alibaba Cloud

Qwen3 from 2

Qwen3.5 from 4

Qwen3 Max from 10

Qwen3.5 Plus from 7

Google

Gemini 3.0 Flash from 4

Gemini 2.5 Pro from 10

Gemini 3.5 Flash from 12

Gemini 3.1 Pro from 15

Gemini 3.0 Flash Search from 6

Gemini 3.5 Flash Search from 18

xAI

Grok 4.3 xAI from 5

Grok 4.3 Search xAI from 25

Perplexity

Perplexity from 2

Perplexity Pro from 20

Image generation

Flux

Flux Schnell

Generation 2

Flux Ultra

Generation 12

Flux 1.1 Pro

Generation 10

Flux Kontext Max

Generation 30

Photo editing 30

Ideogram

Ideogram Turbo 4

Generation 4

Ideogram 3 Turbo

Generation 4

Ideogram 4

Generation 8

Ideogram 3

Generation 8

ByteDance

Seedream 5.0 Lite

Generation 14

Seedream 4.0

Generation 12

Alibaba Cloud

Qwen Lora

	1K	2K	4K
Generation	1	2	8

4K 4:3/3:4 = 6

Google

Google Nano Banana

Generation 15

Photo editing 15

Merge 2 photos 20

Merge 3 photos 25

Merge 4 photos 30

Google Nano Banana 2

	1K	2K	4K
Generation	30	45	60
Photo editing	30	45	60
Merge 2 photos	40	55	70
Merge 3 photos	50	65	80
Merge 4 photos	60	75	90

Google Nano Banana Pro

	1K / 2K	4K
Generation	60	120
Photo editing	60	120
Merge 2 photos	70	130
Merge 3 photos	80	140
Merge 4 photos	90	150

OpenAI

GPT Image 1

Generation 15

Photo editing 20

Merge 2 photos 25

Merge 3 photos 30

Merge 4 photos 35

GPT Image 2

	1K	2K	4K
Generation	20	40	50
Photo editing	25	45	55
Merge 2 photos	30	50	60
Merge 3 photos	35	55	65
Merge 4 photos	40	60	70

2K 16:9/9:16 = 1K

GPT Image 1.5

Generation 12

Photo editing 17

Merge 2 photos 22

Merge 3 photos 27

Merge 4 photos 32

GPT Image 2 High

	1K	2K	4K
Generation	75	150	180
Photo editing	85	160	190
Merge 2 photos	95	170	200
Merge 3 photos	105	180	210
Merge 4 photos	115	190	220

2K 16:9/9:16 = 1K

GPT Image 1.5 High

Generation 40

Photo editing 50

Merge 2 photos 60

Merge 3 photos 70

Merge 4 photos 80

GPT Image 1 High

Generation 50

Photo editing 60

Merge 2 photos 70

Merge 3 photos 80

Merge 4 photos 90

Recraft

Recraft v3

Generation 20

Midjourney

Midjourney 6.1

Generation 20

Midjourney 8.1

Generation 20

Midjourney 7

Generation 20

xAI

Grok Imagine

Generation 10

Grok Imagine Quality

	1K	2K
Generation	25	35
Photo editing	30	40
Merge 2 photos	35	45
Merge 3 photos	40	50

Microsoft

Microsoft MAI-Image-2.5

Generation 25

Photo editing 25

Merge 2 photos 25

Image recognition 10

Video generation

Kling

KLING 1.6 Standard

5 sec 60

10 sec 120

KLING O1

Generation

5 сек 165

10 сек 330

Editing

3 сек 165

4 сек 220

5 сек 275

6 сек 330

7 сек 385

8 сек 440

9 сек 495

10 сек 550

KLING 2.5 Turbo Pro

5 sec 140

10 sec 280

KLING 2.6 Pro

5 sec + sound 280

10 sec + sound 560

KLING 3.0 Standard

5 sec + sound 300

10 sec + sound 600

KLING 2.0 Master

5 sec 560

KLING 3.0 Pro

5 sec + sound 400

10 sec + sound 800

ByteDance

Seedance Lite

5 sec 80

10 sec 160

Seedance 2.0 Fast

5 sec + sound 500

10 sec + sound 1000

Seedance 2.0

5 sec + sound 675

10 sec + sound 1350

Seedance Pro

5 sec 250

10 sec 500

xAI

Grok Imagine Video

5 sec + sound 125

10 sec + sound 250

Google

Gemini Omni

4 sec + sound 455

6 sec + sound 520

8 sec + sound 585

10 sec + sound 650

OpenAI

Sora 2

4 sec 160

8 sec 320

12 sec 480

Sora 2 Pro

4 sec 800

8 sec 1600

12 sec 2400

Veo

Veo 3.1 Fast

8 sec + sound 480

Veo 3.1

8 sec + sound 1250

Max. context length

GPT-4.1 nano / GPT-4.1 / Gemini 3.0 Flash / Gemini 2.5 Pro / Gemini 3.5 Flash / GPT-5.4 / GPT-5.5 / GPT-5.5 High / GPT-5.4 High / Gemini 3.1 Pro / GPT-5.4 Pro / Gemini 3.0 Flash Search / Gemini 3.5 Flash Search 2 000 000

Qwen3.5 Plus 1 000 000

GPT-5.4 nano / GPT-5.4 mini / GPT-5.2 / GPT-5.1 / GPT-5.2 High / GPT-5.1 High 700 000

Grok 4.3 xAI / Grok 4.3 Search xAI 500 000

Claude Haiku 4.5 / Claude Sonnet 4.5 / Claude Sonnet 4.6 / Claude Sonnet High 4.5 / Claude Sonnet High 4.6 / Claude Fable 5 / Claude Opus 4.8 / Claude Opus 4.7 / Claude Opus 4.6 / Claude Opus 4.5 / Claude Opus 4.1 / o3-pro / Perplexity Pro 400 000

Qwen3 / Qwen3.5 / Qwen3 Max / GPT-4o Search 250 000

Perplexity 200 000

DeepSeek-V3.2 160 000

DeepSeek-R1 120 000