Eldrim × the BEAM community
100,000 tokens.
Bring an idea.
Free Swedish AI compute for developers at Stockholm BEAM and Code BEAM 2026. One API key. Open models. No payment card.
Preparing the launch. The GPU and model still need to pass our readiness checks. New grants remain closed until then. The claim page shows current availability.
Planned allocations: 50 Stockholm grants and 300 Code BEAM grants. One grant shared across both events.
A small amount of ceremony. Then code.
- Choose your event. Request a sign-in link to your email address. Public event codes identify the campaign; they are not private invitations.
- Verify and claim. Confirm the link, read the offer and claim your grant. A second claim opens your existing allowance.
- Create a key. Use the console’s Create key and Copy key controls. Save it securely; it is shown once. Lost it? Revoke it and create another on the same grant.
- Run your first request. The console shows used, reserved and available tokens. All keys on the grant share that allowance.
Already claimed with another sign-in? Email sign-in can recover the grant for the same address. Your other Eldrim accounts remain separate.
One working request
Once your grant is active, use https://api.eldrim.net/v1. The campaign model is eldrim-default. Its approved Swedish route will be enabled after provisioning and model checks; it is not a promise of a particular model or throughput.
In Bash, enter the key without saving it in your shell history:
read -rsp 'Eldrim API key: ' ELDRIM_API_KEY; echo
export ELDRIM_API_KEY
curl --fail-with-body https://api.eldrim.net/v1/models \
-H "Authorization: Bearer $ELDRIM_API_KEY"
curl --fail-with-body https://api.eldrim.net/v1/chat/completions \
-H "Authorization: Bearer $ELDRIM_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"eldrim-default","max_tokens":256,"messages":[{"role":"user","content":"Write a small Erlang gen_server and explain its state."}]}'
The model list must include eldrim-default before you send work. If it is empty, check the campaign status in your console. Support: support@eldrim.ai.
Python
Install the SDK in your project’s virtual environment: python -m pip install openai==3.22.1. The example disables automatic retries because repeating a request can consume more tokens.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.eldrim.net/v1",
api_key=os.environ["ELDRIM_API_KEY"],
max_retries=0,
timeout=60,
)
response = client.chat.completions.create(
model="eldrim-default",
max_tokens=256,
messages=[{"role": "user", "content": "Explain Erlang supervision briefly."}],
)
print(response.choices[0].message.content)
Stream an answer
stream = client.chat.completions.create(
model="eldrim-default",
max_tokens=256,
messages=[{"role": "user", "content": "Explain Erlang supervision briefly."}],
stream=True,
stream_options={"include_usage": True},
)
with stream:
for chunk in stream:
if chunk.choices:
print(chunk.choices[0].delta.content or "", end="", flush=True)
if chunk.usage:
print("\nTokens:", chunk.usage.total_tokens)
Streaming returns incremental chat chunks, with text in delta. The final usage chunk can have no choices. See the official OpenAI streaming documentation for the SDK pattern; Eldrim’s supported parameters are listed below.
Erlang
If you already have Eldrim’s Erlang client from the repository, this uses the same endpoint. The client is not published to Hex; contact Eldrim for repository access.
{ok, C} = eldrim_client:new(#{api_key => {env, "ELDRIM_API_KEY"}}),
{ok, Models, _} = eldrim_client:models(C),
{ok, Answer, Meta} = eldrim_client:chat(C, #{
model => <<"eldrim-default">>,
max_tokens => 256,
messages => [#{role => user, content => <<"Explain supervision briefly.">>}]
}).
The offer, precisely
- 100,000 input plus output tokens per verified email identity across the BEAM 2026 campaign family. Replacing a key or attending both events does not reset the allowance. Multiple email addresses are not a way to claim extra grants.
- Ends 6 November 2026 at 23:59 CET. The cutoff for starting work is 7 November at 00:00 in Europe/Stockholm, or
2026-11-06T23:00:00Z. Already admitted, bounded work may finish. - Two concurrent requests and 30 admissions per rolling minute across the grant. Shared capacity also has a global limit; a busy service may refuse a request before dispatch.
- Text chat and streaming only. Use system, user and assistant text messages, one answer and a bounded
max_tokens. Tools, images, audio, Responses, Messages, embeddings, batch jobs and System One are outside this campaign. The broader service has a different compatibility matrix. - Tokens are reserved before work starts. We reserve a conservative input ceiling plus your output cap, then settle reported usage. Missing usage or an interrupted request can leave a reservation pending review. A small remaining balance may not cover that input ceiling.
- A free evaluation grant. It has no cash value, guaranteed availability, automatic renewal or automatic paid overflow. Paid continuation requires a separate agreement. Claims may pause while capacity is unavailable or abuse is investigated.
- Use synthetic or non-sensitive workloads. Regulated and sensitive production uses require separate review. The service terms and privacy notice apply. Grant administration and usage notices do not subscribe you to marketing.
Stockholm code: STOCKHOLM-BEAM. Code BEAM code: CODEBEAM26. Allocation is finite: at most 350 grants and 35 million granted tokens in this family. A full or closed event offers a clear response on the claim page.
Built something worth keeping?
Tell us what you built, its expected load and where it needs to run. We can work out a way to continue when the free grant ends.