Build on Modal at Junction 2026

Get your credits

VVX-2QX-8HW

1. Create a free Modal account at modal.com/signup (GitHub login). One redemption per workspace.
2. Open the redemption link (or scan the QR) and submit your workspace name and the code.
3. Credits appear on modal.com/settings/usage within a minute. Install the client with pip install modal and run modal setup.

Redeem $100 on Modal

One redemption per workspace. Redeem by 16 November 2026. Credits last well beyond the event. Teams: create one workspace, redeem there, invite teammates.

If something goes wrong

• Code already redeemed: each workspace can redeem a code once; check Settings -> Usage first.
• Wrong workspace name: use the workspace slug shown in the dashboard URL, not your email.
• Code expired: ask the organizer or the Modal contact; expiry gates redemption only, credits last well beyond the event.
• GPU job pending: try another GPU type (gpu="A100" or "L40S") or lower the concurrency.

Start here

Install: pip install modal && modal setup. Then clone modal-examples and run any of these, picked for Junction 2026.

modal run 06_gpu_and_ml/llm-serving/vllm_inference.py

modal run 13_sandboxes/codelangchain/agent.py

modal run 13_sandboxes/harbor_evals.py

modal run 06_gpu_and_ml/text-to-audio/chatterbox_tts.py

modal run 06_gpu_and_ml/stable_diffusion/flux.py

modal run 06_gpu_and_ml/paint_flowers/paint_flowers.py

modal run 13_sandboxes/codelangchain/langserve.py

Ways to use Modal this weekend

• Run your agent's tool calls and generated code inside Modal Sandboxes instead of on your laptop
• Serve an open-weight LLM behind a URL with vLLM or SGLang, billed per second, scaling to zero between demos
• Spin up isolated containers per user or per task for code execution, evals and browser automation
• Transcribe or synthesise speech on GPUs (Whisper, Parakeet, TTS) with streaming WebSocket endpoints
• Generate or analyse images and video (Flux, YOLO, VLMs) with weights cached in a Volume

Hardware

Credits work on every GPU Modal offers (T4, L4, A10, L40S, A100, H100, H200, B200, B300) and on CPUs. Pick the GPU in code with gpu="H100" and fall back to another type if a job sits pending; no single GPU type is guaranteed at a given moment. New workspaces start on the Starter plan with its concurrency limits; organizers can ask Modal to raise them for the weekend.

How we expect Modal to fit this event

• GPU-heavy use: expect H100/A100 demand; new workspaces start on Starter limits, so ask Modal to raise concurrency for the weekend and tell teams to fall back to A10/L40S when a type is pending.
• Agent and code-execution projects: Sandboxes are the fit; budget is CPU-dominated, so credits go a long way.
• Real-time audio or video: use WebSocket or WebRTC endpoints and keep containers warm with min_containers during demos.
• Mostly first-time Modal users: run the kick-off workshop early, point teams to hello_world and the examples before they design around a GPU.
• Large event (1200 participants): expect a redemption spike at kick-off; the code has enough uses and the Fillout link is prefilled, so redemption takes under a minute.

Relevant links

• Sandboxes for agents
• Build a coding agent
• LLM inference on Modal
• GPU guide
• Sandbox snapshots
• Whisper transcription
• Streaming and WebSockets
• Image generation with Flux
• Model weights in Volumes
• Modal docs
• Examples repo
• Pricing and GPUs
• Community Slack

The Modal challenge

Best use of Modal at Junction 2026. Use Modal to power the project you are already building: serve a model on a GPU, run your agent's code in Sandboxes, fine-tune on your own data or fan out a batch job. We look for working systems where Modal does the heavy lifting: effective use of Modal (40%), technical ambition (25%), impact (20%), demo quality (15%).

Workshop: Modal workshop, Saturday 14 Nov 13:30, workshop stage

About this event

Junction 2026, Europe's largest hackathon, 13 to 15 November at Hype Areena, Espoo. Theme: European Tech Renaissance. About 1,200 participants, mostly students and young engineers from across Europe, in teams of 2 to 5. Most teams build AI products: LLM apps on open-weight models, agents that run code, voice assistants, computer vision, some fine-tuning. Modal is Technology Partner: booth, Best use of Modal tech challenge, 30 minute workshop on Saturday, social and job posts. Every participant gets $100 in Modal credits; GPUs from T4 to B300; Sandboxes for agent code; Dojo for fine-tuning and RL. Mentors from Modal engineering and DevRel at the booth and on Discord. Prize to be determined.

For organizers: brand kit

• Write "Modal", never "Modal Labs", "MODAL" or "modal" in prose. The company is Modal, the product is Modal.
• Use the supplied logo files; do not recolour, rotate, outline or add effects. Keep clear space of at least the height of the M around it.
• On dark backgrounds use the light logo, on light backgrounds the dark logo. Green accent is
• Do not combine the Modal logo with other sponsor logos into one lockup; place it with equal size and spacing.

Approved blurb (50 words)

Modal is a serverless cloud platform for AI and data teams. Run Python functions on GPUs, serve models as autoscaling endpoints and run agents in secure Sandboxes, billed per second.

Luma description block

Compute partner: Modal. Every participant gets $100 in Modal credits for GPUs and CPUs. Serve models, run agents in Sandboxes, fine-tune, all in Python. Redeem at the link in the participant guide.

Social

We are teaming up with @modal for Junction 2026: every builder gets $100 in GPU credits. Ship an AI demo with a real URL this weekend.

Help

Modal docs, the Modal Community Slack and the examples repo.

Contact: Filip Makraduli; organizer: Emil Pekkinen. Docs: modal.com/docs, community: Modal Slack.

Do credits expire?

The code must be redeemed before its expiry date; the credits themselves last long after the event.

Can a team pool credits?

Credits belong to a workspace. Create one team workspace, redeem once there, and invite teammates (Settings -> Members).

Which GPUs can we use?

All of them, chosen in code with gpu="...". If a type is pending, pick another; no single type is guaranteed at a given moment.

Where do we get help?

Modal docs (modal.com/docs), the Modal Community Slack, the examples repo, and the contacts in this guide.