Elara V5 Pro holds your whole repository in mind: 2,000,000 relevance-ranked tokens, double the window of the frontier chat leaders, anchored with minimised drift. The mathematics is solved, not guessed, and we prove every claim. Start with a 14-day trial.
You should never have to shrink your thinking to fit the tool.
Most coding AIs forget the parts of your repo they cannot see, then fill the gap with a confident guess. Elara holds the whole thing at once and stays anchored to what it knows. The maths brings order to the chaos, so the answer holds up. Below: what it costs, and the proof it works.
14-day trial. Working in Cursor or VS Code in under a minute.
Ask in plain words and Elara does real work: it hosts live rooms where your whole team talks to one Elara, runs your own agents, reads and answers your email, makes images and video, writes documents you can edit, remembers what matters, and reaches you on Telegram. All in one place, no API key needed to start.
Invite anyone into your chat with one link. They join in one press, no account needed, and you both talk to the same Elara with the same memory. Group voice included. The chat leaders retired group chat; ours is the product.
Start a room →Build your own agent or use Hermes, the built-in one. It searches the web, runs code, reads your inbox and works through a plan, and you can watch it work, step by step, and see when it is done.
Meet your agents →Connect your mailbox and Elara reads your inbox, finds the message you mean, reads the attachments, and drafts the reply for you to send. Your whole inbox, triaged in seconds.
Connect your mailbox →Describe it and Elara draws it: photo-real 4K images and cinematic video from one line, right in the chat. One free image every day, for everyone.
Create something →Elara writes formatted documents in a Canvas beside the chat (plans, letters, specs) and you edit them in place, the way you would in a real editor.
Open a canvas →Every agent keeps its own knowledge base, a linked, searchable memory of what it has learned about your work, so it gets sharper every week, and you can read the whole map.
See its memory →Set up your own Telegram bot in three steps and chat to your Elara agent straight from your phone, exactly like here, no app to install.
Set up Telegram →2,000,000-token context holds all of it at once. Your chats, your documents, your agents' knowledge. Elara keeps the whole picture in mind while it works. Try it in the chat →
A bigger window only helps if the model still remembers the middle. We planted a fact deep inside a 2,000,000-token conversation, tried our hardest to make the test unfair, and Elara returned it word for word, every time. Then it kept following the thread out to 8,000,000 tokens, flat. Two independent proof engines checked the result. This is Context Intelligence: memory that stays anchored to what it actually knows, so it does not wander and does not make things up.
The thinking is solved. The only limit is knowledge.Elara holds your whole problem in mind and checks its own work in closed-form mathematics, not a trained guess. Minimised drift, minimised hallucination. While the field is still working to make reasoning trustworthy, that part is finished here, and proven. The only thing that bounds an answer now is what it knows, never whether it wandered, forgot, or made something up.
This is why the AI in your editor is worth paying for: it reads your whole repo, recalls it exactly, and keeps your thread. .LEKOLA-Relevance-Rank
Every plan gives you the full 2,000,000-token window with minimised drift. Pick a flat monthly plan for everyday work, or build on the API and pay for the tokens you use. Every plan starts with a 14-day trial.
.LEKOLA-MAX.LEKOLA-MAXEvery model tier delivers the full 2,000,000-token window with minimised drift, relevance-ranked by the Elara-Cortex. USD per 1,000,000 tokens.
| Tier | Models | Input / 1M | Output / 1M | Note |
|---|---|---|---|---|
| Standard | elara-v5-pro, elara-v4-pro | $4.50 | $22.50 | everyday chat & coding |
| Reasoning | elara-v5-pro-r, elara-v5-deep | $4.50 | $22.50 | checks its own work |
| Ultra | elara-v5-ultra | $4.50 | $22.50 | hardest multi-step work |
.LEKOLA-MAX | full 2,000,000 context + maximum reasoning | $5.00 | $25.00 | maximum power tier, double the industry window |
| Elara coding models · Naledi · Joy · DiGo | ||||
| Naledi | fast, lightweight coding assistant | $0.90 | $4.50 | quick edits & everyday code |
| Joy | balanced coding & reasoning | $2.70 | $13.50 | production-grade work |
| DiGo | deep reasoning & Debug mode | $4.50 | $22.50 | the hardest bugs, highest effort |
Twice the memory, a fraction of the frontier price. Every tier runs the full 2,000,000-token window, double the memory of the frontier chat leaders, at $4.50 in and $22.50 out per million tokens. The .LEKOLA-MAX flagship is $5 and $25. Members save a further 10% on all usage from their monthly allowance, and step up the moment they want more. Your usage is always live in your account.
Elara coding models: Naledi, Joy and DiGo. Three coding assistants for your editor, from quick everyday edits to deep debugging. Naledi is fast and light, Joy is the balanced everyday coder, and DiGo brings deep reasoning with a Debug mode for the hardest problems. Drop them into VS Code or Cursor and pay per token at the rates above. See connect your editor.
Describe it and Elara draws it: photo-real 4K images and cinematic video from a single prompt, made on the Elara models. Studio-grade quality, at a fraction of the frontier price.
Studio quality, a fraction of the price. The quality of the big-name image tools, from under a cent an image. Everyone gets a free image every day; members get video and clean, full-resolution media. Open the studio →
Already signed in? Manage keys, usage & billing →
Send it as a normal Bearer token (OpenAI standard), or as the X-API-Key header.
curl https://www.elara-cortex.com/elara-route/v1/chat/completions \
-H "Authorization: Bearer YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"model":".LEKOLA-MAX","messages":[{"role":"user","content":"Hello"}]}'
Open the chat → Manage keys & usage →
Full API-key management, create, rotate or revoke keys and see usage and credits, lives in your account.
Nothing else changes. If your tool works with OpenAI, it works with Elara: change the base URL and the key.
Base URL: https://www.elara-cortex.com/elara-route/v1 API key: your Elara key (Bearer token)
from openai import OpenAI
client = OpenAI(base_url="https://www.elara-cortex.com/elara-route/v1", api_key="YOUR_KEY")
r = client.chat.completions.create(model=".LEKOLA-MAX",
messages=[{"role":"user","content":"Write a Python function to merge intervals."}])
print(r.choices[0].message.content)
Settings → Models → switch on Override OpenAI Base URL, set the base URL above and your key, then add a model such as .LEKOLA-MAX.
| Model | Context (effective) | Best for |
|---|---|---|
| .LEKOLA-MAX | 2,000,000 | Maximum power, full window |
| .LEKOLA-ACADEMIC | 2,000,000 | Scholarly writing, grounded in live sources |
| elara-v5-ultra | 2,000,000 | Hardest reasoning |
| elara-v5-deep | 2,000,000 | Deep reasoning, long context |
| elara-v5-pro | 2,000,000 | Fast everyday chat & coding |
| Method | Path | Does |
|---|---|---|
| GET | /v1/models | List your models |
| POST | /v1/chat/completions | Chat, streaming or not |
| POST | /v1/responses | Responses-API shape |
Every model has the full 2,000,000-token effective context window, relevance-ranked by the Elara-Cortex; output up to 32K tokens.
| What frustrates people elsewhere | What you get here |
|---|---|
| Usage caps that shrink without warning, exhausted by midweek | You always see what you have left. A live usage counter on every plan, and a free lane that works before you even sign in. |
| Ads on the free tier | No ads. On any tier. |
| Free-tier chats used for training | Your chats stay private. |
| Context that runs out mid-project | A 2,000,000-token effective window on every model, so whole codebases and long projects fit. |
| Answers you have to double-check yourself | Verification runs on the live path, and room answers carry a receipt with real reasoning seconds. |
| Working alone in a single-user chat | Live rooms. Invite anyone into the same chat, voice included, no account needed to join. Neither of the big two has this. |
These claims are probed against the live product on a standing 37-point parity battery, and where a rival is better today we say so in the benchmarks. Honest even when we lose.
The proof is public. Read the benchmarks, read the drift methodology, or just open the chat and drop in your hardest problem.