Add Grok 4.20 reasoning and non-reasoning to xAI model catalog by Jaaneek · Pull Request #50772 · openclaw/openclaw

Jaaneek · 2026-03-20T02:39:53Z

Summary

Problem: Grok 4.20 is now GA but only the experimental beta entries (grok-4.20-experimental-beta-0304-*) exist in the catalog.
Why it matters: Users can't select the GA model IDs from the bundled catalog or model picker.
What changed: Added grok-4.20-reasoning and grok-4.20-non-reasoning to the xAI catalog, removed the superseded beta entries, updated docs and tests.
What did NOT change: Forward-compatibility resolver, web search defaults, onboarding flow, auth.

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

Related: xAI Grok 4.20 GA release

User-visible / Behavior Changes

xai/grok-4.20-reasoning and xai/grok-4.20-non-reasoning now appear in openclaw models list and can be set as the default model.
grok-4.20-experimental-beta-0304-reasoning and grok-4.20-experimental-beta-0304-non-reasoning are removed from the bundled catalog (still forward-resolved by the grok-4.20 prefix matcher).

Security Impact (required)

New permissions/capabilities? No
Secrets/tokens handling changed? No
New/changed network calls? No
Command/tool execution surface changed? No
Data access scope changed? No

Repro + Verification

Environment

OS: macOS 24.3.0
Runtime/container: Node 22.22.0
Model/provider: xai

Steps

Start onboarding
Select grok 4.20 reasoning
Send messages to test

Expected

Both models visible in models list with yes auth status
Model resolves and responds

Actual

Both models appear, API confirms grok-4.20-0309-reasoning resolves correctly

Evidence

Failing test/log before + passing after
pnpm test -- extensions/xai/provider-models.test.ts — 5/5 passed
pnpm check — all passed
Direct API test: curl https://api.x.ai/v1/chat/completions with grok-4.20-reasoning returns valid response

Human Verification (required)

Verified scenarios: models list, direct API call, TUI model picker
Edge cases checked: no edge cases afaik
What you did not verify: Selecting the model when another model was already selected from previous version, I only tried on brand new openclaw with new onboarding

Review Conversations

I replied to or resolved every bot review conversation I addressed in this PR.
I left unresolved only the conversations that still need reviewer or maintainer judgment.

Compatibility / Migration

Backward compatible? Yes
Config/env changes? No
Migration needed? No

Failure Recovery (if this breaks)

How to disable/revert this change quickly: Revert commit
Files/config to restore: extensions/xai/model-definitions.ts
Known bad symptoms reviewers should watch for: None expected

Risks and Mitigations

None

greptile-apps · 2026-03-20T02:41:35Z

Greptile Summary

This PR adds grok-4.20-reasoning and grok-4.20-non-reasoning to the xAI model catalog, each backed by the existing 2M-token XAI_LARGE_CONTEXT_WINDOW constant, along with updated documentation and a new test case that validates both entries. The implementation is clean and follows the established pattern for catalog entries.

model-definitions.ts: Two new XaiCatalogEntry objects added; type, flags, and constant usage are all correct.
provider-models.test.ts: New test covers both model IDs with the expected reasoning flag and context window.
docs/providers/xai.md: Catalog list updated; models are ordered (stable before experimental) consistently with typical convention.
Minor ordering inconsistency: The docs list the new stable models before the grok-4.20-experimental-beta-0304-* variants, but the catalog in model-definitions.ts places them after. Aligning the code order to match the docs would improve consistency.

Confidence Score: 5/5

This PR is safe to merge; changes are additive, well-tested, and follow the existing catalog pattern exactly.
The diff is minimal and low-risk: two new catalog entries, one new test block, and a one-line doc update. No existing behaviour is altered, the new entries use already-validated constants, and the fallback resolution logic in resolveXaiCatalogEntry was already correct for the grok-4.20 prefix. The only note is a cosmetic ordering inconsistency between docs and code.
No files require special attention.

Prompt To Fix All With AI

This is a comment left during a code review.
Path: extensions/xai/model-definitions.ts
Line: 73-84

Comment:
**Catalog ordering differs from docs**

The new `grok-4.20-reasoning` / `grok-4.20-non-reasoning` entries are placed **after** the `grok-4.20-experimental-beta-0304-*` entries in the catalog, but the docs (`docs/providers/xai.md`) list them **before** the experimental beta variants. Since the stable models are conventionally preferred over the experimental ones, aligning the catalog order with the docs would make both easier to reason about:

```suggestion
  {
    id: "grok-4.20-reasoning",
    name: "Grok 4.20 (Reasoning)",
    reasoning: true,
    contextWindow: XAI_LARGE_CONTEXT_WINDOW,
  },
  {
    id: "grok-4.20-non-reasoning",
    name: "Grok 4.20 (Non-Reasoning)",
    reasoning: false,
    contextWindow: XAI_LARGE_CONTEXT_WINDOW,
  },
  {
    id: "grok-4.20-experimental-beta-0304-reasoning",
    name: "Grok 4.20 Experimental Beta 0304 (Reasoning)",
    reasoning: true,
    contextWindow: XAI_LARGE_CONTEXT_WINDOW,
  },
  {
    id: "grok-4.20-experimental-beta-0304-non-reasoning",
    name: "Grok 4.20 Experimental Beta 0304 (Non-Reasoning)",
    reasoning: false,
    contextWindow: XAI_LARGE_CONTEXT_WINDOW,
  },
```

This is a cosmetic point and has no functional impact, but it keeps the code and docs in sync.

How can I resolve this? If you propose a fix, please make it concise.

_{Last reviewed commit: "xai: add grok-4.20-r..."}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 363edc0cb5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-20T17:52:49Z

+  if (provider === "xai") {
+    return normalizeXaiModelId(model);


Preserve legacy xAI keys when normalizing model refs

This new normalization rewrites xai/grok-4.20-experimental-beta-0304-* to the GA IDs before selection, but resolveThinkingDefault still reads cfg.agents.defaults.models by the normalized key and does not check the deprecated xAI key form. In an upgraded config that still stores per-model params.thinking under the old key, that override is now silently ignored and the runtime falls back to global/default thinking. Please add a legacy-key fallback (or migrate keys) so backward compatibility also preserves per-model settings.

Useful? React with 👍 / 👎.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e2fff01389

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-20T18:49:29Z

+  if (provider === "xai") {
+    return normalizeXaiModelId(model);


Keep legacy xAI model keys when reading per-model params

Normalizing deprecated grok-4.20-experimental-beta-0304-* IDs at model-ref parse time changes the selected key to xai/grok-4.20-*, but downstream lookups still read cfg.agents.defaults.models by exact provider/model (for example in resolveExtraParams and resolveFastModeState). In upgraded configs that still store params under the old xAI keys, settings like temperature, maxTokens, and fastMode are silently ignored, so the backward-compat path is incomplete unless we also fall back to the legacy key (or migrate it).

Useful? React with 👍 / 👎.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fa3919b83f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-20T19:27:32Z

@@ -121,6 +121,9 @@ function normalizeProviderModelId(provider: string, model: string): string {
  if (provider === "google" || provider === "google-vertex") {
    return normalizeGoogleModelId(model);
  }
+  if (provider === "xai") {
+    return normalizeXaiModelId(model);


Normalize legacy xAI IDs before unique-provider inference

parseModelRef now rewrites legacy Grok 4.20 beta IDs to GA IDs, but inferUniqueProviderFromConfiguredModels still compares the caller-supplied model string verbatim against parsed allowlist entries. With a legacy session/runtime model like grok-4.20-experimental-beta-0304-reasoning (no modelProvider) and an allowlist containing that same legacy key, the allowlist side is normalized to grok-4.20-reasoning while the runtime side is not, so provider inference fails and downstream identity/cost paths in resolveSessionModelIdentityRef lose the provider unexpectedly.

Useful? React with 👍 / 👎.

Grok 4.20 is now GA. This adds both variants to the bundled catalog and removes the superseded experimental beta entries that shipped during the preview period. - Add grok-4.20-reasoning (2M context, reasoning enabled) - Add grok-4.20-non-reasoning (2M context, reasoning disabled) - Remove grok-4.20-experimental-beta-0304-reasoning - Remove grok-4.20-experimental-beta-0304-non-reasoning - Update provider docs and tests

huntharo · 2026-03-20T19:28:34Z

Merged via squash.

Prepared head SHA: 095e645ea58b2259b25c923aeaf11bbcb2990c8f
Merge commit: 916f496b5190c34e8514a11b260ea23409775872

Thanks @Jaaneek!

@huntharo

…law#50772) Merged via squash. Prepared head SHA: 095e645 Co-authored-by: Jaaneek <25470423+Jaaneek@users.noreply.github.com> Co-authored-by: huntharo <5617868+huntharo@users.noreply.github.com> Reviewed-by: @huntharo

@huntharo

…law#50772) Merged via squash. Prepared head SHA: 095e645 Co-authored-by: Jaaneek <25470423+Jaaneek@users.noreply.github.com> Co-authored-by: huntharo <5617868+huntharo@users.noreply.github.com> Reviewed-by: @huntharo

@huntharo

…law#50772) Merged via squash. Prepared head SHA: 095e645 Co-authored-by: Jaaneek <25470423+Jaaneek@users.noreply.github.com> Co-authored-by: huntharo <5617868+huntharo@users.noreply.github.com> Reviewed-by: @huntharo

@huntharo

…law#50772) Merged via squash. Prepared head SHA: 095e645 Co-authored-by: Jaaneek <25470423+Jaaneek@users.noreply.github.com> Co-authored-by: huntharo <5617868+huntharo@users.noreply.github.com> Reviewed-by: @huntharo

openclaw-barnacle Bot added docs Improvements or additions to documentation size: XS labels Mar 20, 2026

greptile-apps Bot reviewed Mar 20, 2026

View reviewed changes

Comment thread extensions/xai/model-definitions.ts

Jaaneek force-pushed the feat/add-grok-4.20-models branch from e7535e9 to 7e26c0b Compare March 20, 2026 03:22

Jaaneek changed the title ~~xai: add grok-4.20-reasoning and grok-4.20-non-reasoning models~~ Add Grok 4.20 reasoning and non-reasoning to xAI model catalog Mar 20, 2026

huntharo self-assigned this Mar 20, 2026

openclaw-barnacle Bot added gateway Gateway runtime agents Agent runtime and tooling size: S and removed size: XS labels Mar 20, 2026

chatgpt-codex-connector Bot reviewed Mar 20, 2026

View reviewed changes

huntharo force-pushed the feat/add-grok-4.20-models branch from df9b0fc to e2fff01 Compare March 20, 2026 18:43

chatgpt-codex-connector Bot reviewed Mar 20, 2026

View reviewed changes

huntharo force-pushed the feat/add-grok-4.20-models branch 3 times, most recently from 00b1b16 to fa3919b Compare March 20, 2026 19:21

chatgpt-codex-connector Bot reviewed Mar 20, 2026

View reviewed changes

huntharo force-pushed the feat/add-grok-4.20-models branch from fa3919b to b322ea1 Compare March 20, 2026 19:27

Jaaneek and others added 5 commits March 20, 2026 15:28

xAI: normalize deprecated Grok 4.20 ids

73079d1

xAI: normalize reply-time stored overrides

6dc39b1

xAI: document Grok 4.20 GA ids

52fc2f4

xAI: place Grok 4.20 changelog entry at section tail

095e645

huntharo force-pushed the feat/add-grok-4.20-models branch from b322ea1 to 095e645 Compare March 20, 2026 19:28

huntharo merged commit 916f496 into openclaw:main Mar 20, 2026
10 checks passed

github-actions Bot mentioned this pull request Mar 20, 2026

📡 Upstream Digest — 2026-03-20 20:23 UTC curtismercier/openclaw-mods#318

Open

clawsweeper Bot mentioned this pull request May 3, 2026

Support for the new Grok 4.2 Multi Model #38204

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add Grok 4.20 reasoning and non-reasoning to xAI model catalog#50772

Add Grok 4.20 reasoning and non-reasoning to xAI model catalog#50772
huntharo merged 5 commits intoopenclaw:mainfrom
Jaaneek:feat/add-grok-4.20-models

Jaaneek commented Mar 20, 2026 •

edited

Loading

Uh oh!

greptile-apps Bot commented Mar 20, 2026

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Mar 20, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Mar 20, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Mar 20, 2026

Uh oh!

Uh oh!

huntharo commented Mar 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

Jaaneek commented Mar 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

Security Impact (required)

Repro + Verification

Environment

Steps

Expected

Actual

Evidence

Human Verification (required)

Review Conversations

Compatibility / Migration

Failure Recovery (if this breaks)

Risks and Mitigations

Uh oh!

greptile-apps Bot commented Mar 20, 2026

Greptile Summary

Confidence Score: 5/5

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

huntharo commented Mar 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Jaaneek commented Mar 20, 2026 •

edited

Loading