Modal Changelog
Product updates and improvements to Modal.
-
Inkling 2 Small is now available on Modal
Inkling 2 Small is the latest open-weight model from Thinking Machines. Built for agentic coding, visual reasoning, and long-running research tasks, it supports text and image input and a 1M-token context window.
-
New Sandbox Events UI for execs, TCP tunnels, and HTTP requests
Using the new Sandbox Events UI, you can see what's happening inside a sandbox, right in your dashboard.
-
View invoices directly in the Modal dashboard
Workspace managers can now view all historical invoices in the new Invoices tab under Usage & Billing. Check invoice status and line items, download PDFs, and filter to unpaid invoices without leaving the dashboard.
-
Kimi K3.1 Pro is now available on Shared Endpoints
Access Moonshot AI’s Kimi K3.1 Pro through a Shared Endpoint with per-token billing on Modal. A major post-training update to Kimi K3, it supports coding, financial research, and long-horizon agentic work, with text and image input and a 1M-token context window.
-
New Image Builder version 2026.10 with updated trust stores
We’ve released Image Builder version 2026.10, which includes updates such as refreshed trust stores following the recent eMudhra CA compromise and improvements to build caching.
-
DeepSeek V4.1 Flash 1017 is now available on Shared Endpoints
Access DeepSeek V4.1 Flash 1017 through a Shared Endpoint with per-token billing on Modal. This model is an updated checkpoint of DeepSeek V4.1 Flash, which improves coding, reasoning, and agentic capabilities; it includes support for text and image input and a 1M-token context window.
-
Exit Snapshots for Sandboxes available in Beta
Exit Snapshots take a filesystem snapshot automatically on both graceful and ungraceful Sandbox exits, including terminate(), entrypoint completion, idle timeout, lifetime expiry, and OOM.
-
Sunsetting Modal Notebooks, effective today
As of October 23rd, Modal Notebooks are no longer available. Users have changed the ways they work with Modal, with many preferring the IDE or agent harness of their choice. Existing notebooks can be downloaded indefinitely as IPython notebook files. If you have any questions or concerns, please contact support@modal.com.
-
New Logs Search Experience
App logs and audit logs have a new search bar experience. You type your query as plain text, so you can edit filters with the keyboard like any other text instead of clicking through dropdowns. New operators are also available within app logs.
-
Reflection Beam is now available on Modal
Deploy Reflection’s Beam as a Dedicated Endpoint on Modal. Built for coding, reasoning, and agentic workloads, Beam has 501B total parameters, 23B active per token, and a 1M-token context window.
-
Muse Spark Open is now available on Modal
Access Meta’s Muse Spark Open through a Shared Endpoint with per-token billing on Modal. Built for agentic coding and long-horizon tasks, Muse Spark Open supports multimodal understanding and a 1M-token context window.
-
NVIDIA Nemotron 3.5 Super VL is now available on Modal
Deploy NVIDIA’s Nemotron 3.5 Super VL as a Dedicated Endpoint on Modal. With 120B total parameters and 12B active per token, it supports visual reasoning across text, images, and video for document understanding and agentic workflows.
-
Step 5 Preview is now available on Modal
Deploy StepFun’s Step 5 Preview as a Dedicated Endpoint on Modal. This multimodal model supports software engineering and long-horizon agentic tasks, with 600B total parameters, 27B active per token, and a 1M-token context window.
-
Faster grep on Volumes v2 for Sandboxes
We've recently made grep faster on Volumes v2 (~10x on cold caches and ~6x on warm caches).
-
Give agents a full computer with VM Sandboxes
VM Sandboxes give your agent access to a full Linux computer.
-
Faster Containers and Storage pages
Filtering, sorting, and navigation are now faster on the Containers and Storage pages.
-
Sandbox Sidecars, a new isolation primitive, are now in Beta
Sidecars are containers that run alongside your main Sandbox on the same host, but with the same isolation as separate sandboxes.
-
Python SDK 1.6.0 and JavaScript and Go SDKs 0.11.0
We’ve released version 1.6.0 of the Python SDK and version 0.11.0 of the JavaScript and Go SDKs.
-
Set expiration times for API and service-user tokens
You can now set a token lifetime when creating API tokens and service-user tokens in the dashboard. Workspace managers can also enforce maximum lifetimes for member tokens created in their workspace.
-
Network egress usage broken down by app
The Usage page now shows network egress per app.