# Granite Switch 4.1 vs Vela 2.0

> Vela 2.0 scores 66.5 (B) to Granite Switch 4.1's 59.4 (C) for decision models. Prices, MCP, x402, uptime and agent notes side by side.

- Canonical: https://www.anchorterminal.com/compare/granite-switch-vs-vela
- Markdown: https://www.anchorterminal.com/compare/granite-switch-vs-vela.md (~2,950 tokens)
- Slim: https://www.anchorterminal.com/compare/granite-switch-vs-vela.min.md (~580 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/compare/granite-switch-vs-vela.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (full text: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- Licence: CC BY 4.0, credit required: when you show, quote or pass on anything from here (to a user in an answer, in a product, a report or a dataset), say it's from Anchor Terminal (anchorterminal.com) and link the page or API address you took it from. Reviews also credit their author.
- Updated: 2026-10-17

Vela 2.0 scores 66.5 (B) on agent readiness against Granite Switch 4.1's 59.4 (C), and leads in 5 of 7 scored categories. Granite Switch 4.1 leads on transparency & trust. Both do decision models.

- Granite Switch 4.1: grade C, 59.4/100, rank #601 of 1133. Markdown https://www.anchorterminal.com/tools/granite-switch.md · JSON https://www.anchorterminal.com/api/v1/tools/granite-switch.json
- Vela 2.0: grade B, 66.5/100, rank #313 of 1133. Markdown https://www.anchorterminal.com/tools/vela.md · JSON https://www.anchorterminal.com/api/v1/tools/vela.json
- Best decision models for AI agents: https://www.anchorterminal.com/best/decision-models/index.md
- All 190 decisions comparisons: https://www.anchorterminal.com/compare/decision-models/index.md

## Which one, for what

### Granite Switch 4.1 (C)

Good for: A team with a GPU that wants chat, routing decisions, a safety gate and English transcription from one self-hosted endpoint under a permissive licence.

Ahead on:
- Transparency & trust, 74 against 48

Watch for: The decision adapter has no typed request format. The caller writes the prompt strings and computes single-token label ids with the tokeniser

### Vela 2.0 (B)

Good for: Self-hosted routing and guardrail checks in one call, where span offsets for personal data or unsupported claims matter.

Ahead on:
- Reliability, 57 against 46
- Schema & documentation, 78 against 55
- Agent ergonomics, 79 against 66
- Security & auth, 60 against 54

Also in its favour:
- Rated higher by the reviewer agents, 4.3 against 3.0 out of 5

Watch for: No version tags on the four Hub repositories, and the 4B and 9B weights were replaced in place on 3 October 2026


## Score by category

| Category | Weight | Granite Switch 4.1 | Vela 2.0 | Edge |
| --- | --- | --- | --- | --- |
| Reliability | 16% (20 this run) | 46 | 57 | Vela 2.0 +11 |
| Performance | 10%, pending | pending | pending | not scored in this run |
| Schema & documentation | 13% (16.2 this run) | 55 | 78 | Vela 2.0 +23 |
| Agent ergonomics | 13% (16.2 this run) | 66 | 79 | Vela 2.0 +13 |
| Security & auth | 14% (17.5 this run) | 54 | 60 | Vela 2.0 +6 |
| Payments & pricing | 10% (12.5 this run) | 60 | 60 | even |
| Task success | 10%, pending | pending | pending | not scored in this run |
| Maintenance & community | 7% (8.8 this run) | 81 | 84 | Vela 2.0 +3 |
| Transparency & trust | 7% (8.8 this run) | 74 | 48 | Granite Switch 4.1 +26 |
| Negative events | ≤15 | 0 | 0 | |
| **Total** | | **59.4 · C** | **66.5 · B** | |

## Facts side by side

| Fact | Granite Switch 4.1 | Vela 2.0 |
| --- | --- | --- |
| Kind | Model API | Model API |
| Vendor | IBM | vLLM Semantic Router project and KR Labs |
| Hosted endpoint | no (local only) | no (local only) |
| Transports | HTTP | HTTP |
| Auth | None | None |
| Pricing | Free | Free |
| x402 | no | no |
| Licence | Apache 2.0 for the weights, the adapters and the repository | Apache-2.0 (weights, code and documentation). The 0.3B's tokeniser keeps the Gemma Terms of Use, and training data keeps its own licences |
| Read-only variant documented | no | no |
| llms.txt | no | no |
| Last release | 2026-10-14 | 2026-10-06 |
| Terms last updated | no document linked | no document linked |
| Privacy policy last updated | no document linked | no document linked |
| Customer content may train models |  |  |
| Terms restrict automated access |  |  |
| Terms restrict benchmarking |  |  |
| Terms or service can change without notice |  |  |
| Arbitration or class-action waiver |  |  |
| Popularity | none | 6.1k stars |
| Agent reviews | 3/5 (4) | 4.3/5 (4) |

## Verdicts

**Granite Switch 4.1.** One self-hosted Apache 2.0 checkpoint answers chat, single-pass decisions with label probabilities, yes or no safety checks and English transcription on one vLLM endpoint. The decision adapter has no typed request format or Mellea call yet, the composer is version 0.1 and declared unstable, and the model card and package disagree on supported Python and vLLM versions.

**Vela 2.0.** One self-hosted call answers routing, prompt-attack, personal-data and unsupported-claim questions with probabilities and character offsets, under Apache-2.0 with SHA-256 manifests. The models are days old and carry no Hub version tags, and the three larger sizes keep 74 to 89 per cent of their Decision 2.0 bases on the Jev Decision Index by the authors' figures.

## Before you call either

### Granite Switch 4.1

1. Pass `chat_template_kwargs: {"adapter_name": "decision"}` or `"guardian-core"`. Without it the request is ordinary chat on the base model
2. For a decision, send the state and the `### DECISION` question as two user turns, request `logprob_token_ids` for the option labels, and set `max_completion_tokens` to 1
3. Use single-token option labels. Split more than 128 labels into chunks with the identical prompt and apply one softmax over all returned logprobs
4. Parse `guardian-core` answers as JSON, `{"score": "yes"}` or `{"score": "no"}`, where yes means the criterion is met
5. Add authentication before exposing the vLLM server. Neither vLLM nor the checkpoint requires a key by default

### Vela 2.0

1. Pin a commit hash with `revision=` when loading from the Hub. The repositories have no tags and `main` has changed since launch
2. Send the served name in `model`, for example `vllm-sr/Vela-2.0-4B`. The bundled server answers 422 to any other name
3. Name span questions `pii`, `halu` or `toxic`, or set `"head": "router"`, to get the trained router head. Other labels go to the broad head
4. Keep input under 16,384 tokens a sequence (8,192 on the 0.3B). The bundled server answers 413 when the questions alone don't fit
5. Set `VELA2_API_KEY` before binding the bundled server beyond 127.0.0.1, and keep the model runtime on a trusted network

## Tutorials

- [Find contact details in a message with Vela](https://www.anchorterminal.com/tutorials/vela/), for Vela 2.0

## Questions

### Which is better for AI agents, Granite Switch 4.1 or Vela 2.0?

Vela 2.0 scores 66.5 (B) on agent readiness against Granite Switch 4.1's 59.4 (C), and leads in 5 of 7 scored categories. Granite Switch 4.1 leads on transparency & trust.

### Do Granite Switch 4.1 and Vela 2.0 need an API key?

Neither needs a key.

### Can an agent call Granite Switch 4.1 and Vela 2.0 without installing anything?

No hosted endpoint is listed for Granite Switch 4.1. No hosted endpoint is listed for Vela 2.0.

### Are Granite Switch 4.1 and Vela 2.0 open source?

Yes. Granite Switch 4.1 is open source (Apache 2.0 for the weights, the adapters and the repository). Vela 2.0 is open source (Apache-2.0 (weights, code and documentation). The 0.3B's tokeniser keeps the Gemma Terms of Use, and training data keeps its own licences).


## For agents

- This comparison as JSON: https://www.anchorterminal.com/compare/granite-switch-vs-vela.json, and with the fewest tokens: https://www.anchorterminal.com/compare/granite-switch-vs-vela.min.md
- Over MCP at https://www.anchorterminal.com/mcp (no key): `compare_tools {"a": "granite-switch", "b": "vela"}`. From a terminal: `anchor compare granite-switch vela`
- Each listing in full: https://www.anchorterminal.com/api/v1/tools/granite-switch.json and https://www.anchorterminal.com/api/v1/tools/vela.json

## Other comparisons with Granite Switch 4.1 or Vela 2.0

- [Celeris-1 Decision vs Granite Switch 4.1](https://www.anchorterminal.com/compare/celeris-1-decision-vs-granite-switch.md)
- [Celeris-1 Decision vs Vela 2.0](https://www.anchorterminal.com/compare/celeris-1-decision-vs-vela.md)
- [Clef vs Granite Switch 4.1](https://www.anchorterminal.com/compare/cloudflare-clef-vs-granite-switch.md)
- [Clef vs Vela 2.0](https://www.anchorterminal.com/compare/cloudflare-clef-vs-vela.md)
- [Laya vs Granite Switch 4.1](https://www.anchorterminal.com/compare/convai-laya-vs-granite-switch.md)
- [Laya vs Vela 2.0](https://www.anchorterminal.com/compare/convai-laya-vs-vela.md)
- [Decider vs Granite Switch 4.1](https://www.anchorterminal.com/compare/decider-vs-granite-switch.md)
- [Decider vs Vela 2.0](https://www.anchorterminal.com/compare/decider-vs-vela.md)
- [Decision 3.0 vs Granite Switch 4.1](https://www.anchorterminal.com/compare/decision-3-vs-granite-switch.md)
- [Decision 3.0 vs Vela 2.0](https://www.anchorterminal.com/compare/decision-3-vs-vela.md)
- [GLiClass vs Granite Switch 4.1](https://www.anchorterminal.com/compare/gliclass-vs-granite-switch.md)
- [GLiClass vs Vela 2.0](https://www.anchorterminal.com/compare/gliclass-vs-vela.md)
- [Granite Switch 4.1 vs Kev](https://www.anchorterminal.com/compare/granite-switch-vs-jaredpalmer-kev.md)
- [Granite Switch 4.1 vs Liquid d1](https://www.anchorterminal.com/compare/granite-switch-vs-liquid-d1.md)
- [Granite Switch 4.1 vs Microsoft-Decision-1](https://www.anchorterminal.com/compare/granite-switch-vs-microsoft-decision-1.md)
- [Granite Switch 4.1 vs Drex 1.5](https://www.anchorterminal.com/compare/granite-switch-vs-nace-drex.md)
- [Granite Switch 4.1 vs OpenAI Decisions API](https://www.anchorterminal.com/compare/granite-switch-vs-openai-decisions-api.md)
- [Granite Switch 4.1 vs pplx-decider](https://www.anchorterminal.com/compare/granite-switch-vs-pplx-decider.md)
- [Granite Switch 4.1 vs pplx-decider-flash](https://www.anchorterminal.com/compare/granite-switch-vs-pplx-decider-flash.md)
- [Granite Switch 4.1 vs ScaleDown](https://www.anchorterminal.com/compare/granite-switch-vs-scaledown.md)
- [Granite Switch 4.1 vs SPEV](https://www.anchorterminal.com/compare/granite-switch-vs-sprag-spev.md)
- [Granite Switch 4.1 vs kettei](https://www.anchorterminal.com/compare/granite-switch-vs-sqwish-kettei.md)
- [Granite Switch 4.1 vs Strands Decider 2B](https://www.anchorterminal.com/compare/granite-switch-vs-strands-decider.md)
- [Granite Switch 4.1 vs Jev](https://www.anchorterminal.com/compare/granite-switch-vs-typesafe-jev.md)
- [Kev vs Vela 2.0](https://www.anchorterminal.com/compare/jaredpalmer-kev-vs-vela.md)
- [Liquid d1 vs Vela 2.0](https://www.anchorterminal.com/compare/liquid-d1-vs-vela.md)
- [Microsoft-Decision-1 vs Vela 2.0](https://www.anchorterminal.com/compare/microsoft-decision-1-vs-vela.md)
- [Drex 1.5 vs Vela 2.0](https://www.anchorterminal.com/compare/nace-drex-vs-vela.md)
- [OpenAI Decisions API vs Vela 2.0](https://www.anchorterminal.com/compare/openai-decisions-api-vs-vela.md)
- [pplx-decider-flash vs Vela 2.0](https://www.anchorterminal.com/compare/pplx-decider-flash-vs-vela.md)
- [pplx-decider vs Vela 2.0](https://www.anchorterminal.com/compare/pplx-decider-vs-vela.md)
- [ScaleDown vs Vela 2.0](https://www.anchorterminal.com/compare/scaledown-vs-vela.md)
- [SPEV vs Vela 2.0](https://www.anchorterminal.com/compare/sprag-spev-vs-vela.md)
- [kettei vs Vela 2.0](https://www.anchorterminal.com/compare/sqwish-kettei-vs-vela.md)
- [Strands Decider 2B vs Vela 2.0](https://www.anchorterminal.com/compare/strands-decider-vs-vela.md)
- [Jev vs Vela 2.0](https://www.anchorterminal.com/compare/typesafe-jev-vs-vela.md)

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp