# InstaVM vs Together Code Sandbox (slim)

> Together Code Sandbox scores 54.7 (C) to InstaVM's 52.1 (D) for code sandboxes. Prices, MCP, x402, uptime and agent notes side by side.

- Full: https://www.anchorterminal.com/compare/instavm-vs-together-code-sandbox.md (~4,900 tokens) · this version ~2,530 tokens · JSON https://www.anchorterminal.com/compare/instavm-vs-together-code-sandbox.json · canonical https://www.anchorterminal.com/compare/instavm-vs-together-code-sandbox
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-07
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

Together Code Sandbox scores 54.7 (C) on agent readiness against InstaVM's 52.1 (D), and leads in 5 of 7 scored categories. InstaVM leads on reliability. Both do code sandboxes.

- InstaVM: D 52.1, rank #994 of 1486 · https://www.anchorterminal.com/tools/instavm.min.md
- Together Code Sandbox: C 54.7, rank #896 of 1486 · https://www.anchorterminal.com/tools/together-code-sandbox.min.md
- InstaVM, good for Hosted microVM sandboxes with browser, desktop, OCI images, clones and public shares. Ahead on Reliability, 77 against 32.
- Together Code Sandbox, good for Teams already buying from Together, and former CodeSandbox SDK users, who want Docker-defined sandboxes with disk and memory snapshots from Python or TypeScript. Ahead on Schema & documentation, 86 against 66; Agent ergonomics, 70 against 51; Security & auth, 46 against 41; Maintenance & community, 83 against 35.

| Category | InstaVM | Together Code Sandbox |
| --- | --- | --- |
| Reliability (16%) | 77 | 32 |
| Performance (10%) | pending | pending |
| Schema & documentation (13%) | 66 | 86 |
| Agent ergonomics (13%) | 51 | 70 |
| Security & auth (14%) | 41 | 46 |
| Payments & pricing (10%) | 20 | 20 |
| Task success (10%) | pending | pending |
| Maintenance & community (7%) | 35 | 83 |
| Transparency & trust (7%) | 57 | 59 |

| Fact (where they differ) | InstaVM | Together Code Sandbox |
| --- | --- | --- |
| Vendor | Checksum Labs, Inc. | Together AI |
| Hosted endpoint | `https://api.instavm.io` | `https://api.bartender.codesandbox.io` |
| Licence | none | Proprietary service under Together's terms of service. The SDKs, CLI and OpenAPI documents on GitHub are MIT |
| Last release | 2026-07-15 | 2026-10-07 |
| Terms last updated | 2026-01-26 | no date given |
| Privacy policy last updated | 2026-01-26 | no date given |
| Customer content may train models | yes | not found in the text |
| Terms restrict automated access | yes | not found in the text |
| Terms restrict benchmarking | not found in the text | yes |
| Terms or service can change without notice | yes | not found in the text |
| Arbitration or class-action waiver | yes | not found in the text |
| Popularity | 57 npm/wk, 370 PyPI/wk | 3 stars, 1.3k npm/wk |

Why the edge:
- Maintenance & community, Together Code Sandbox +48: InstaVM 35 (Closed service lines. The latest dated change found is Python SDK 0.26.1 on 15 July 2026, 114 days before the check, and the site changelog's only entry is 11 July (10). No SDK release in the 90 days to 6 November (0). A public changelog, support email and Discord, with no reply times seen (8 of 15). Official Python and JavaScript SDKs, last released in July and May 2026 (12 of 15). Package health of the SDK repositories was not read (5 of 10). Total 35.); Together Code Sandbox 83 (Last release `together-sandbox` 4.1.1 on 7 October 2026, 28 days before 4 November (30). Four tagged releases from 6 August, 4.0.3 (10 August), 4.0.4 (30 September), 4.1.0 (1 October) and 4.1.1 (7 October), each changing SDK or API code (20). The repository's open issues and pull requests weren't readable on 4 November (github.com refuses us), so the 8 October reading stands, when the 30 most recent items were team pull requests merged within days (12). Current official Python and TypeScript SDKs and a CLI, versioned together (15). Pull requests run the TypeScript unit tests, end-to-end tests run on pushes to main, and packages publish by OIDC trusted publishing. The Python tests are not in the pull request workflow, and run results weren't read (6). 30 + 20 + 12 + 15 + 6 = 83.)
- Reliability, InstaVM +45: InstaVM 77 (Hosted lines. Better Stack status page with four components, including the public API (20). The incident history shows no incidents for September, October or November 2026 (30). No rate limits with numbers found; Pro promises higher limits without figures (0). The Python SDK has a RateLimitError and an example retry loop with exponential backoff, but no 429 or Retry-After is documented, and there is no idempotency guidance (7 of 15). The terms commit to 99.9% monthly uptime with service credits, and Pro lists 99.9% (10). No beta or preview label on the API (10). Total 77.); Together Code Sandbox 32 (Read with the hosted lines. status.together.ai lists the website, the playground and inference models with no component for the sandbox service, and a vendor status page with no component for the product takes 15. Incident line kept at 5. The Together page posts no incident since 6 January 2025, while status.codesandbox.io, which lists the CodeSandbox API and two clusters, wasn't re-read on 4 November (codesandbox.io answers a bot check, so its terms are unread), per the 8 October check (5). No rate or concurrency limits found, and the migration guide says the concurrency count and limit are not exposed (0). Both SDKs retry 408, 429, 500, 502, 503 and 504 three times with exponential backoff from 0.5 seconds plus jitter, and the SDK references say `snapshots.create` is not idempotent and should be excluded from retries. That is safe-retry guidance with no idempotency keys (12). No SLA found, and the terms make no uptime guarantee outside an order form (0). The docs still say the SDK must be enabled for an organisation, on request to sales, which is not general availability (0). 15 + 5 + 0 + 12 + 0 + 0 = 32.)
- Schema & documentation, Together Code Sandbox +20: InstaVM 66 (A public OpenAPI 3.1 spec with 229 operations (25). llms.txt at instavm.io (10). Only 93 of 229 operations have a description, and those read as short summaries without when-not-to-use guidance (8 of 20). FastAPI-generated request schemas with typed fields and required lists (10 of 15). Docs carry code examples for each SDK area, but only 422 validation errors appear in the spec and no error-code reference was found (6 of 15). Paths sit under /v1 and an SDK changelog exists, but the site changelog has one entry, of 11 July 2026 (7 of 15). Total 66.); Together Code Sandbox 86 (Two public OpenAPI documents in the repository and on GitHub Pages, the management API (3.1.0, 17 operations) and the in-VM API (3.0.3, 25 operations) (25). docs.together.ai has llms.txt and a Markdown twin of the page, and each package ships an `LLMS.md` with the references (10). The concept docs say when to snapshot memory rather than disk only and what ephemeral means, while 36 of 114 typed nodes in the management document carry a description (14). Enums for status, bounds on `cpu`, `memory_bytes`, `ttl` and `limit`, with `experimental` left loose (12). Examples in Python, TypeScript and the CLI, and an error object with `code`, `message` and `errors`. No list of error codes found, and 429 is not in the document (10). `/v1` paths, semver packages and a generated `CHANGELOG.md` (15).)
- Agent ergonomics, Together Code Sandbox +19: InstaVM 51 (30 operations take limit, offset, cursor or page parameters, with no field selection found (12 of 25). Listing endpoints page, with little filtering (10 of 20). The SDK maps failures to typed exceptions such as AuthenticationError, QuotaExceededError and RateLimitError, but the HTTP error bodies are undocumented (8 of 20). One operation takes an idempotency parameter, and snapshots and clones make work repeatable, with no safe-retry guidance (6 of 20). `client.execute(code=..., language=...)` needs two arguments, with Python and JavaScript SDKs (15). Total 51.); Together Code Sandbox 70 (List calls take `limit` from 1 to 100, batch reads take IDs and exec output can be read from a `lastSequence`. No field selection (15). Cursor pagination on sandboxes and snapshots, with filters for status, source snapshot, tags and retired snapshots (20). Errors return `code`, `message` and per-parameter entries, the SDKs raise one `HttpError` with the status, and `status_reason` names causes such as `out_of_capacity` and `oom_killed`. No catalogue of codes found (14). Built-in retries, but no idempotency keys, sandbox IDs can't be chosen and `snapshots.create` can register duplicates (8). Python and TypeScript SDKs and a CLI, with defaults of 1 vCPU and 2 GiB. A snapshot must be built first because no default image exists (13).)
- Security & auth, Together Code Sandbox +5: InstaVM 41 (Plain API keys in a header, with API endpoints to create, list and delete them and no scopes found (20 of 30). Egress policies per session or VM can block all outbound traffic or allow named domains and CIDRs, but llms.txt describes full network access as the default (8 of 20). No guidance on untrusted output from executed code or browser sessions (3 of 15). An Audit Logs endpoint group exists in the API reference (10 of 15). No security.txt, disclosure policy, bug bounty or certification found; the privacy policy names none (0 of 20). Total 41.); Together Code Sandbox 46 (Project-scoped API keys, revocable within about five minutes, with optional expiry and rotation guidance, read as the Together inference and fine-tuning listings read the same key model (30). The in-VM OpenAPI document still lists `?token=` as a documented way to authenticate the WebSocket exec path, so the checklist's fixed 10 comes off (20). Each sandbox is a virtual machine. Every port is open to everyone unless the experimental inbound policy is set, no outbound rules exist, and project roles are admin and editor with no read-only role (8). The sandbox returns the output of code the agent ran, with no guidance on untrusted output found (8). Sandboxes can be listed with tags and `status_reason`. No audit log found (3). A PGP-signed security.txt (Expires 2028-09-30) points Contact and Policy at HackerOne with no acknowledgement time, the value this run gives a disclosure route with no stated time (7). 20 + 8 + 8 + 3 + 7 = 46.)