Six text models on Kumo can be used for free: Claude Fable 5.2, Claude Opus 5.5, GPT-6.1 Sol, Grok 4.7, DeepSeek V4.1 Flash and GLM 5.3 Flash. GPT Image 2.6 Sunburst covers images. Requests cost nothing in chat and by API key for as long as they fit the limit. Below: how to get access, how much it gives and at what point a request becomes paid.
Free models
6 + 1
six text models and one for images
Start
Free
a Kumo channel subscription on Telegram and a verified email
Daily limit
from $3
with a top-up or a package, resets at 00:00 MSK
Which models are free
The list as of 12 November 2026. These are the same models as in the paid catalogue: the same gateway, the same protocols, the same key. There is one difference: while a request fits your free limit, nothing is charged for it.
| Model | Vendor | What it is good for |
|---|---|---|
| Claude Fable 5.2 | Anthropic | Demanding reasoning and long agent work |
| Claude Opus 5.5 | Anthropic | Everyday coding and analysis |
| GPT-6.1 Sol | OpenAI | A general model for text and code |
| Grok 4.7 | xAI | Conversation and code |
| DeepSeek V4.1 Flash | DeepSeek | A low-cost model for many requests |
| GLM 5.3 Flash | Zhipu AI | A fast, low-cost model |
| GPT Image 2.6 Sunburst | OpenAI | Image generation |
The cards below show the normal prices of these models. That is what a request costs when it does not fit the free limit.
Anthropic
Claude Fable 5.2
anthropic/claude-fable-5-2
Anthropic's model for demanding reasoning and long-horizon agentic work. 1M-token context, up to 128K output tokens. Available over the Anthropic Messages and Chat Completions protocols.
- Input
- 5.7 USD≈ 486 ₽
- per 1M tokens
- Output
- 28.5 USD≈ 2429 ₽
- per 1M tokens
Context: 1M
Pricing and access on Kumo →Anthropic
Claude Opus 5.5
anthropic/claude-opus-5-5
Anthropic Claude Opus 5.5 text model. Available over the Anthropic Messages and Chat Completions protocols.
- Input
- 2.2 USD≈ 187 ₽
- per 1M tokens
- Output
- 11 USD≈ 937 ₽
- per 1M tokens
Context: 1M
Pricing and access on Kumo →OpenAI
GPT-6.1 Sol
openai/gpt-6.1-sol
OpenAI GPT-6.1 Sol text model. Available over the Chat Completions and Responses protocols.
- Input
- 1.1 USD≈ 94 ₽
- per 1M tokens
- Output
- 5.5 USD≈ 469 ₽
- per 1M tokens
Context: 1.1M
Pricing and access on Kumo →xAI
Grok 4.7
xai/grok-4.7
xAI Grok 4.7 text model. Available over the Chat Completions and Responses protocols.
- Input
- 1.1 USD≈ 94 ₽
- per 1M tokens
- Output
- 3.3 USD≈ 281 ₽
- per 1M tokens
Context: 500K
Pricing and access on Kumo →DeepSeek
DeepSeek V4.1 Flash
deepseek-v4-1-flash
The economy tier of DeepSeek's V4.1 family: a sparse mixture-of-experts model on a 552B-parameter backbone that activates about 8B parameters for input and 16B for output. Image understanding is trained in from the start, and a compressed KV cache lowers the cost of long agent sessions. It fits agents that write code, drive a terminal or operate a computer, as well as long-document analysis, with a 1M-token context and tool calling.
- Input
- 0.1 USD≈ 10 ₽
- per 1M tokens
- Output
- 0.5 USD≈ 41 ₽
- per 1M tokens
Context: 1M
Pricing and access on Kumo →Zhipu
GLM 5.3 Flash
glm-5.3-flash
A fast, low-cost Zhipu GLM-5.3 model and the first natively multimodal one in the line: it reads text, images and video. Suited to coding, visual interface work and documents, with a 1M-token context, reasoning and tool calling.
- Input
- 0.2 USD≈ 15 ₽
- per 1M tokens
- Output
- 0.6 USD≈ 50 ₽
- per 1M tokens
Context: 1M
Pricing and access on Kumo →OpenAI
GPT Image 2.6 Sunburst
openai/gpt-image-2.6-sunburst
OpenAI GPT Image 2.6 Sunburst image-generation model. Available over the Images protocol.
- Image
- 0.06 USD≈ 5 ₽
- per image
How to get access
You do not have to pay to start. You need a Kumo account and Telegram.
- Sign up in the Kumo console and confirm your email with the link from the message.
- Subscribe to the Kumo news channel on Telegram. The console shows the link on the free models card.
- In the same place link Telegram to your account: sign in with Telegram, or enter your username and press Start in the bot. Kumo checks the subscription itself.
- Open the model catalogue, switch on the «Free» filter and pick a model marked Free. Or send a request by API with your ordinary key.
After that the account gets the first step, Intro.
Intro: a one-time bonus to try the models
The first step gives a one-time bonus instead of a daily limit: about $0.8 at the models' prices. Each model spends it at its own price. So on the low-cost DeepSeek V4.1 Flash the bonus covers roughly 1,322 typical requests, on Claude Opus 5.5 roughly 51, and on Claude Fable 5.2 about 15.
The bonus is given once per Kumo account and once per Telegram account. It lasts 30 days from your first free request and does not renew daily. While the bonus is active, you can create one image a week for free.
The estimate assumes a typical request: 12 thousand input tokens, 85% of them read from the cache, and 800 output tokens. The console counts it the same way. Short chat questions spend the bonus more slowly, large requests faster.
Once the bonus is spent or its term is over, free requests stop. They come back when the account gets a daily step: a top-up or a package.
The daily limit: top-ups and packages
A lasting free limit comes with money on the account. It renews every day at 00:00 Moscow time and grows from step to step.
- Top-up. It needs the channel subscription, a balance of $3 or more and a top-up of at least that amount within the last 30 days. The step has four levels: from $3, from $20, from $50 and from $100. The level follows the current balance: as the balance is spent, the level goes down.
- Packages Starter, Pro, Max and Ultra. A package gives a bigger limit than a top-up of the same amount. Without the channel subscription a package gives 80% of its limit.
How many requests a day a step gives
| Step | Images a week | DeepSeek V4.1 Flash | GPT-6.1 Sol | Claude Opus 5.5 | Claude Fable 5.2 |
|---|---|---|---|---|---|
| Top-up from $3 | 1 | ≈355 | ≈24 | ≈8 | 0 |
| Top-up from $20 | 1 | ≈435 | ≈31 | ≈11 | 0 |
| Top-up from $50 | 2 | ≈677 | ≈53 | ≈22 | ≈4 |
| Top-up from $100 | 5 | ≈899 | ≈73 | ≈32 | ≈8 |
| Starter | 2 | ≈556 | ≈42 | ≈17 | ≈1 |
| Pro | 3 | ≈878 | ≈71 | ≈31 | ≈7 |
| Max | 5 | ≈1,362 | ≈114 | ≈53 | ≈16 |
| Ultra | 10 | ≈1,805 | ≈154 | ≈73 | ≈24 |
Claude Fable 5.2 is the dearest model on the list. For one typical request to it Kumo reserves about $0.3. On the top-up levels from $3 and from $20 the daily limit is smaller than that, so the request does not fit, and on Starter one request fits. Regular free work with Fable 5.2 takes the Pro package or higher, while Intro is enough to try it.
For a daily step to work the account has to keep a reserve: $2 or more on the balance, or 2M tokens or more in the package. Below that the console shows «Free models are paused» and requests go at the normal price.
When a request is free and when it is not
All free text models share one limit. The console shows it as a single «Today's limit» meter: dear models spend it faster, cheap ones more slowly.
Before every request Kumo works out the most it can cost: your input at the dearest input rate plus an answer of the maximum length. If that amount fits what is left of the limit, the request is free. If it does not, the request goes at the normal price, even when part of the limit is still there. That is why by the end of the day a dear model may no longer fit while the cheap ones keep working for free.
What else to know
- The answer of a free request is capped at 8,192 tokens. If you need a longer answer, split the task into parts.
- Two requests run free at the same time. A third one sent in parallel goes at the normal price.
- The daily limit renews at 00:00 Moscow time. What you did not use is not carried over.
- Images are counted separately, by the piece per week. The counter resets on Monday at 00:00 Moscow time.
- The Intro and Top-up steps share a daily promotion budget across all participants. When it is spent, the console says so and requests go at the normal price until midnight.
- Claude Fable 5.2 has its own bound on the size of a free request: roughly up to 141,500 characters of input.
What happens when the limit runs out
A request to a free model that did not fit the limit is handled as an ordinary paid one: its cost is charged to the package or the balance.
If the account has no money, the request does not run: the API answers with error 402. Nothing is charged for it.
Free models by API
No separate key or header is needed. A request to a listed model with your ordinary Kumo key goes through the same limit check as a request from chat. Streaming works too.
curl -s https://api.kumorouter.com/v1/chat/completions \
-H "Authorization: Bearer $KUMO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-opus-5-5",
"messages": [
{"role": "user", "content": "Explain what this function does: ..."}
],
"max_tokens": 2000
}'The model field takes the name of any free model:
claude-fable-5-2
claude-opus-5-5
gpt-6.1-sol
grok-4.7
deepseek-v4-1-flash
glm-5.3-flashThe same works in Cursor, Claude Code and other clients that connect to Kumo with a key. Keep in mind that agents send large requests: the limit is spent faster than in chat, and dear models may not fit it. For agent work on the free limit DeepSeek V4.1 Flash and GLM 5.3 Flash suit better.
Where to see what is left
- Console home: the «Limits» card with the «Today's limit» meter and the count of images for the week.
- Model catalogue: the «Free» filter and the Free mark on a model with an estimate of how many requests to it are still free.
- Request history: free rows carry the price $0.00 with the normal price struck through.
- Activity: the «Free models» table shows how much you saved in 30 days.
The promotion terms in brief
- The promotion is temporary. The list of models and the size of the limits change; the console always shows the current values.
- One Telegram account links to one Kumo account. Several accounts for the sake of extra limit break the terms, and participation in the promotion is switched off for it.
- The free limit cannot be saved up, transferred or exchanged for money or tokens.
- When a top-up is refunded, the cost of the free requests made on the Top-up step is withheld from the refund.
- The full text is in the free models promotion terms.
FAQ
- Do I have to pay to try it?
- No. For the Intro step it is enough to subscribe to the Kumo channel on Telegram and confirm your email. No card and no top-up are needed.
- Will money be charged when the limit runs out?
- Only if the account has some. A request beyond the limit is paid from the package or the balance at the normal price. If the balance is empty, the request does not run and the API answers with error 402.
- Can I use Claude Opus 5.5 for free every day?
- Yes, on the daily steps. As of 12 November 2026 a top-up from $3 gives about 8 typical requests to Opus 5.5 a day, the Pro package about 31, Ultra about 73. Short requests spend the limit more slowly.
- Why does a request to Claude Fable 5.2 not fit the limit?
- The most a request can cost is reserved for it; for a typical request to Fable 5.2 that is about $0.3. On the lower top-up levels the daily limit is smaller than that. Fable 5.2 is free on Intro, on packages and on the top-up levels from $50.
- Do free models work in Cursor and Claude Code?
- Yes, with your ordinary Kumo API key. But agents send large requests, so the limit is spent faster and dear models may not fit it. For agents on the free limit DeepSeek V4.1 Flash and GLM 5.3 Flash suit better.
- Can I get Intro a second time?
- No. The bonus is given once per Kumo account and once per Telegram account.
- When does the limit renew?
- The daily limit of the text models renews every day at 00:00 Moscow time. The image counter resets on Monday at 00:00. What is left is not carried over.
Where to start
The list of models and their normal prices is in the catalogue. You can try the free models right away in the Kumo console: sign up, subscribe to the channel, and the first request goes out at no charge.
