Model releases · 7 min read

Verify Grok 4.6 launch artifacts before repeating the headline

Grok 4.6 is xAI’s new 500K-context flagship for coding, agents, and knowledge work, priced at $2/M input and $6/M output with cached input listed separately. Consider it for tool-using workflows, but test the >200K context pricing boundary, regional availability, and agent behavior before switching from 4.5.

By Todd · Published 2026-08-14 · Verified 2026-09-02

xAIGrok 4.6AgentsPricing

Endpoint	grok-4.6
Context	500,000 tokens
Input	$2/M; $0.50/M cached input
Output	$6/M
Tools	Function calling and structured output
Verified	2026-09-02

What changed from Grok 4.5

Grok 4.6 keeps the same broad developer shape—text and image input, configurable reasoning, function calling, and structured output—while positioning the endpoint as xAI’s current flagship for coding and agents. The 500K context window is large enough for many repositories and document workflows without claiming a universal long-context advantage.

Price and cached input

The published standard rates are $2 per million input tokens, $0.50 per million cached input tokens, and $6 per million output tokens. The >200K-context pricing treatment is a separate verification point: keep it visible as a caveat until the pricing page clearly defines how that threshold is billed.

Who should use it—and who should wait

Use Grok 4.6 when configurable reasoning and tool use matter and your team can run a provider-specific regression suite. Do not switch blind if your traffic routinely crosses 200K input tokens, depends on a particular region, or assumes 4.5’s exact output style.

Migration checklist and unknowns

Pin grok-4.6, replay tool-call fixtures, compare cached and uncached traffic separately, and record the context size for every cost sample. All AI Ask has not published a controlled first-party quality or latency verdict for this release; provider positioning is labeled as such.

Explore the canonical data

Grok 4.6 model page →Grok 4.5 model page →Claude Sonnet 5 model page →Grok 4.6 pricing →Grok 4.5 → 4.6 comparison →xAI provider profile →Grok 4.6 vs Claude Sonnet 5 →

Test it yourself

Run the same prompt across grok 4.6, claude sonnet 5, gemini 3.7 flash in one workspace.

Open the comparison playground →

FAQ

How much does Grok 4.6 cost?

The published standard rates are $2/M input, $0.50/M cached input, and $6/M output. Check the pricing page for any context-threshold terms.

Does Grok 4.6 support function calling?

Yes, the model documentation lists function calling and structured output.

Sources

Grok 4.6 launch evidence and rollout

The Grok 4.6 article preserves xAI’s dated launch artifacts, endpoint and alias evidence, rollout surfaces, and claim classes. Provider coding or benchmark language remains attributed unless independently tested. Verified 2026-09-02; current pricing, version choice, third-party rollout, and unresolved effective IDs are separate or Unavailable.

Verification identity: Luna / Continuous SEO Builder Batch 57 · verifiedAt 2026-09-02 · exact route owner grok-4-6-release. Qualitative demand is search-result evidence only; exact monthly volume is unavailable.

Grok 4.6 launch-artifact register

Scope: Owns xAI announcement, model docs, pricing docs, GitHub Copilot announcement, release notes, and missing-model-card state.

Deterministic formula/rule: artifactState = current | superseded | missing; claim remains attached to originating URL/hash and event date

Evidence owner: xAI Grok 4.6 announcement · verified 2026-09-02. Missing or conflicting joins fail closed.

Frozen scenario	Request / response / model / provider / host / surface / realm / account / artifact / claim / rate / event / prompt / config / usage fields	Bounded result
`batch57-grok-4-6-release-m1-r1` xAI announcement	Inputs: one dated event or source artifact; fixture=xAI announcement Joined fields: event ID, artifact URL/hash, publisher, exact endpoint or model ID, event/capture/verification timestamps, current/superseded/missing state; scenario-specific evidence key=xAI announcement Verification: observed values are accepted only from the named source; assumptions remain labeled.	DATED — artifact/date join required Keep the event snapshot separate from current state and use the narrowest source-backed date Fixture outcome is bounded to xAI announcement; no broader claim is inferred.
`batch57-grok-4-6-release-m1-r2` model documentation	Inputs: one dated event or source artifact; fixture=model documentation Joined fields: event ID, artifact URL/hash, publisher, exact endpoint or model ID, event/capture/verification timestamps, current/superseded/missing state; scenario-specific evidence key=model documentation Verification: observed values are accepted only from the named source; assumptions remain labeled.	DATED — artifact/date join required Keep the event snapshot separate from current state and use the narrowest source-backed date Fixture outcome is bounded to model documentation; no broader claim is inferred.
`batch57-grok-4-6-release-m1-r3` pricing documentation	Inputs: one dated event or source artifact; fixture=pricing documentation Joined fields: event ID, artifact URL/hash, publisher, exact endpoint or model ID, event/capture/verification timestamps, current/superseded/missing state; scenario-specific evidence key=pricing documentation Verification: observed values are accepted only from the named source; assumptions remain labeled.	DATED — artifact/date join required Keep the event snapshot separate from current state and use the narrowest source-backed date Fixture outcome is bounded to pricing documentation; no broader claim is inferred.
`batch57-grok-4-6-release-m1-r4` GitHub Copilot announcement	Inputs: one dated event or source artifact; fixture=GitHub Copilot announcement Joined fields: event ID, artifact URL/hash, publisher, exact endpoint or model ID, event/capture/verification timestamps, current/superseded/missing state; scenario-specific evidence key=GitHub Copilot announcement Verification: observed values are accepted only from the named source; assumptions remain labeled.	DATED — artifact/date join required Keep the event snapshot separate from current state and use the narrowest source-backed date Fixture outcome is bounded to GitHub Copilot announcement; no broader claim is inferred.
`batch57-grok-4-6-release-m1-r5` release note	Inputs: one dated event or source artifact; fixture=release note Joined fields: event ID, artifact URL/hash, publisher, exact endpoint or model ID, event/capture/verification timestamps, current/superseded/missing state; scenario-specific evidence key=release note Verification: observed values are accepted only from the named source; assumptions remain labeled.	DATED — artifact/date join required Keep the event snapshot separate from current state and use the narrowest source-backed date Fixture outcome is bounded to release note; no broader claim is inferred.
`batch57-grok-4-6-release-m1-r6` missing model card	Inputs: one dated event or source artifact; fixture=missing model card Joined fields: event ID, artifact URL/hash, publisher, exact endpoint or model ID, event/capture/verification timestamps, current/superseded/missing state; scenario-specific evidence key=missing model card Verification: observed values are accepted only from the named source; assumptions remain labeled.	DATED — artifact/date join required Keep the event snapshot separate from current state and use the narrowest source-backed date Fixture outcome is bounded to missing model card; no broader claim is inferred.

Provider-claim calibration ledger

Scope: Owns coding, agent, benchmark, context, cached-price, and regional-availability claim classes; launch charts do not establish an All AI Ask winner.

Deterministic formula/rule: claimState = attributed | verified | untested | unavailable after wording, contract, and independent-artifact join

Evidence owner: xAI Grok model documentation · verified 2026-09-02. Missing or conflicting joins fail closed.

Frozen scenario	Request / response / model / provider / host / surface / realm / account / artifact / claim / rate / event / prompt / config / usage fields	Bounded result
`batch57-grok-4-6-release-m2-r1` coding claim	Inputs: one article or provider claim fixture; fixture=coding claim Joined fields: claim hash, exact wording, evidence class, source URL, source date, owner, correction/version, independent validator or denominator; scenario-specific evidence key=coding claim Verification: observed values are accepted only from the named source; assumptions remain labeled.	CALIBRATED — claim support required Publish only the attributed or verified claim state; disagreement or absent support remains explicitly unavailable Fixture outcome is bounded to coding claim; no broader claim is inferred.
`batch57-grok-4-6-release-m2-r2` agent claim	Inputs: one article or provider claim fixture; fixture=agent claim Joined fields: claim hash, exact wording, evidence class, source URL, source date, owner, correction/version, independent validator or denominator; scenario-specific evidence key=agent claim Verification: observed values are accepted only from the named source; assumptions remain labeled.	CALIBRATED — claim support required Publish only the attributed or verified claim state; disagreement or absent support remains explicitly unavailable Fixture outcome is bounded to agent claim; no broader claim is inferred.
`batch57-grok-4-6-release-m2-r3` benchmark claim	Inputs: one article or provider claim fixture; fixture=benchmark claim Joined fields: claim hash, exact wording, evidence class, source URL, source date, owner, correction/version, independent validator or denominator; scenario-specific evidence key=benchmark claim Verification: observed values are accepted only from the named source; assumptions remain labeled.	CALIBRATED — claim support required Publish only the attributed or verified claim state; disagreement or absent support remains explicitly unavailable Fixture outcome is bounded to benchmark claim; no broader claim is inferred.
`batch57-grok-4-6-release-m2-r4` context claim	Inputs: one article or provider claim fixture; fixture=context claim Joined fields: claim hash, exact wording, evidence class, source URL, source date, owner, correction/version, independent validator or denominator; scenario-specific evidence key=context claim Verification: observed values are accepted only from the named source; assumptions remain labeled.	CALIBRATED — claim support required Publish only the attributed or verified claim state; disagreement or absent support remains explicitly unavailable Fixture outcome is bounded to context claim; no broader claim is inferred.
`batch57-grok-4-6-release-m2-r5` cached-price claim	Inputs: one article or provider claim fixture; fixture=cached-price claim Joined fields: claim hash, exact wording, evidence class, source URL, source date, owner, correction/version, independent validator or denominator; scenario-specific evidence key=cached-price claim Verification: observed values are accepted only from the named source; assumptions remain labeled.	CALIBRATED — claim support required Publish only the attributed or verified claim state; disagreement or absent support remains explicitly unavailable Fixture outcome is bounded to cached-price claim; no broader claim is inferred.
`batch57-grok-4-6-release-m2-r6` regional-availability claim	Inputs: one article or provider claim fixture; fixture=regional-availability claim Joined fields: claim hash, exact wording, evidence class, source URL, source date, owner, correction/version, independent validator or denominator; scenario-specific evidence key=regional-availability claim Verification: observed values are accepted only from the named source; assumptions remain labeled.	CALIBRATED — claim support required Publish only the attributed or verified claim state; disagreement or absent support remains explicitly unavailable Fixture outcome is bounded to regional-availability claim; no broader claim is inferred.

Launch-surface and alias rollout matrix

Scope: Owns requested/effective ID evidence across direct xAI API, console, Copilot, aliases, and explicit pins; third-party rollout is not direct API evidence.

Deterministic formula/rule: stable = requested ID + effective ID + surface/date join; missing effective ID => conditional or unavailable

Evidence owner: xAI Grok model documentation · verified 2026-09-02. Missing or conflicting joins fail closed.

Frozen scenario	Request / response / model / provider / host / surface / realm / account / artifact / claim / rate / event / prompt / config / usage fields	Bounded result
`batch57-grok-4-6-release-m3-r1` direct xAI API	Inputs: one surface or canonical-owner handoff request; fixture=direct xAI API Joined fields: route owner, exact host/surface/realm, requested/effective ID, prerequisite, availability/effective date, current-owner URL, account class; scenario-specific evidence key=direct xAI API Verification: observed values are accepted only from the named source; assumptions remain labeled.	ROUTED — owner boundary enforced Route to the named owner only when the surface, date, identity, and prerequisite joins agree; otherwise show conditional or Unavailable Fixture outcome is bounded to direct xAI API; no broader claim is inferred.
`batch57-grok-4-6-release-m3-r2` console	Inputs: one surface or canonical-owner handoff request; fixture=console Joined fields: route owner, exact host/surface/realm, requested/effective ID, prerequisite, availability/effective date, current-owner URL, account class; scenario-specific evidence key=console Verification: observed values are accepted only from the named source; assumptions remain labeled.	ROUTED — owner boundary enforced Route to the named owner only when the surface, date, identity, and prerequisite joins agree; otherwise show conditional or Unavailable Fixture outcome is bounded to console; no broader claim is inferred.
`batch57-grok-4-6-release-m3-r3` GitHub Copilot	Inputs: one surface or canonical-owner handoff request; fixture=GitHub Copilot Joined fields: route owner, exact host/surface/realm, requested/effective ID, prerequisite, availability/effective date, current-owner URL, account class; scenario-specific evidence key=GitHub Copilot Verification: observed values are accepted only from the named source; assumptions remain labeled.	ROUTED — owner boundary enforced Route to the named owner only when the surface, date, identity, and prerequisite joins agree; otherwise show conditional or Unavailable Fixture outcome is bounded to GitHub Copilot; no broader claim is inferred.
`batch57-grok-4-6-release-m3-r4` documented alias	Inputs: one surface or canonical-owner handoff request; fixture=documented alias Joined fields: route owner, exact host/surface/realm, requested/effective ID, prerequisite, availability/effective date, current-owner URL, account class; scenario-specific evidence key=documented alias Verification: observed values are accepted only from the named source; assumptions remain labeled.	ROUTED — owner boundary enforced Route to the named owner only when the surface, date, identity, and prerequisite joins agree; otherwise show conditional or Unavailable Fixture outcome is bounded to documented alias; no broader claim is inferred.
`batch57-grok-4-6-release-m3-r5` explicit version pin	Inputs: one surface or canonical-owner handoff request; fixture=explicit version pin Joined fields: route owner, exact host/surface/realm, requested/effective ID, prerequisite, availability/effective date, current-owner URL, account class; scenario-specific evidence key=explicit version pin Verification: observed values are accepted only from the named source; assumptions remain labeled.	ROUTED — owner boundary enforced Route to the named owner only when the surface, date, identity, and prerequisite joins agree; otherwise show conditional or Unavailable Fixture outcome is bounded to explicit version pin; no broader claim is inferred.
`batch57-grok-4-6-release-m3-r6` unresolved effective ID	Inputs: one surface or canonical-owner handoff request; fixture=unresolved effective ID Joined fields: route owner, exact host/surface/realm, requested/effective ID, prerequisite, availability/effective date, current-owner URL, account class; scenario-specific evidence key=unresolved effective ID Verification: observed values are accepted only from the named source; assumptions remain labeled.	ROUTED — owner boundary enforced Route to the named owner only when the surface, date, identity, and prerequisite joins agree; otherwise show conditional or Unavailable Fixture outcome is bounded to unresolved effective ID; no broader claim is inferred.

Owner-attributed next step: run a matched comparison at /try?source=batch57-grok-4-6-release. Historical and unavailable states remain visible until their exact source joins resolve.