xAI Models: Every Model, Spec, Price and Release Date

xAI is the developer of the Grok family of large language models, offered through the xAI API; its current flagship, Grok 4.7 (API id grok-4.7), became available on the xAI API on 21 September 2026. This page lists every model on xAI's current pricing and models pages, with API identifiers, context windows, published prices, rate limits and status, plus the version history, retirements and an index of model pages. Every fact comes from xAI's own developer documentation, release notes, model pages or pricing page. The developer documentation site is currently titled "SpaceXSI Docs" and describes Grok 4.7 as "SpaceXSI's frontier model"; it is served from the x.ai domain. Where xAI publishes no value, the page says "Not stated by xAI" rather than estimating. Facts are as of 15 November 2026.

Last checked: 2026-11-15.

This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider.

Key takeaways

• Grok 4.7 is the flagship. API id grok-4.7, a 500k token context window, text and image input, text output, configurable reasoning effort (low, medium, high, xhigh, default high) and a stated knowledge cutoff of May 2026.
• Several Grok 4 versions stay priced side by side. grok-4.7, grok-4.6, grok-4.5, grok-4.3 and three Grok 4.20 variants are all on the pricing page, plus the coding model grok-build-0.1 in early access.
• Prices step up above 200k tokens. Once a prompt reaches 200k tokens, long-context rates apply to all tokens in that request; for grok-4.7 that is $4.00 input and $12.00 output per million instead of $2.00 and $6.00.
• No Grok model on the pricing page is open weight. All listed Grok models are proprietary and served through the xAI API.
• The catalogue goes beyond chat models. xAI also prices Grok Imagine image and video models, Grok Voice speech-to-speech, speech-to-text and text-to-speech, and server-side tools such as web_search and x_search.

What is xAI?

xAI publishes the Grok models through a single API with a global endpoint and a US regional endpoint. Model identifiers follow a documented alias scheme: <modelname> points to the latest stable version, <modelname>-latest to the latest version, and <modelname>-<date> to a fixed release, for example grok-4.20-0309-reasoning. Grok 4.20 is split into reasoning, non-reasoning and multi-agent identifiers, while later versions use one identifier with a reasoning effort parameter. See the LLM Model Naming and Versioning Explained page for how these conventions work across providers.

The models page highlights Grok 4.7, the Grok Voice API and image and video generation, and was last updated on 21 September 2026. The pricing page lists the full set of billable text, image, video and voice models, which is used below as the current catalogue.

Current xAI models at a glance

As of 15 November 2026. Source: xAI pricing page, models page and release notes. Text models are listed newest first by version number, then media models.

Grok 4.7 Fast is named in the release notes as available only in Cursor and Grok Build, not on the public API, so it has no API row.

Key facts for the main xAI models

As of 15 November 2026. Identical fields for every model; "Not stated by xAI" means not stated by xAI on the pages checked (2026-11-15). API IDs, context, licence and status are in the table above.

Grok 4.20 lists text and image modalities without splitting them into input and output. Grok Build 0.1 release is early access on 2026-05-19, after the Grok Build beta launch on 2026-05-14.

API features and limits listed on the model pages

Model pages were checked for Grok 4.7 and Grok 4.20 reasoning; xAI publishes no separate model page text for the other versions on the pages checked. Image input accepts JPG/JPEG and PNG up to 20 MiB per image, according to the models page. See Reasoning Models Explained.

xAI model families and naming

xAI groups its models into the lines below, as of 15 November 2026. Release dates for each are in the next section.

• Grok 4 line: Grok 4 (July 2025), Grok 4.1 Fast (November 2025), Grok 4.20 (March 2026), Grok 4.3 (release date not stated by xAI), Grok 4.5 (July 2026), Grok 4.6 (August 2026) and Grok 4.7 (September 2026). Grok 4 and Grok 4.1 Fast do not appear on the current pricing page; the models page links a "Model Retirement (May 15, 2026)" guide whose details were not available when checked.
• Grok Build: a coding product (beta May 2026) with its own model, grok-build-0.1. Grok 4.7 Fast is offered only in Cursor and Grok Build.
• Grok Imagine: image models (grok-imagine-image, grok-imagine-image-2.0) and video models (grok-imagine-video, grok-imagine-video-1.5, grok-imagine-video-1.5-lite).
• Grok Voice: speech-to-speech models from Voice 1.0 (April 2026) to Voice 3.0 (October 2026), plus transcription models.

When was each xAI model released?

Dates from the xAI release notes, as of 15 November 2026, in date order. The release notes list 2026 entries by month and day under year headings.

Grok 4.7 became available on the xAI API on 21 September 2026. The most recent release note entries are Grok Voice 3.0 (14 October 2026) and the retirement of the Grok Imagine image quality and pro models (2 November 2026).

How much do xAI models cost?

As of 15 November 2026. USD per million tokens on the global endpoint. Source: xAI pricing page (docs.x.ai/developers/pricing). Long-context rates apply to all tokens in a request once its prompt reaches 200k tokens. Notes shows input / cached / output at 200k+. xAI does not state a separate cache write price on the pricing page.

• Batch: the 20% discount applies to input, output, cached and reasoning tokens. Image and video generation are billed at standard rates in the Batch API.
• Priority Processing: billed at 2x standard rates for text requests.
• US regional endpoint: billed at 1.1x global rates and currently available only for grok-4.7 and grok-4.6; for grok-4.7 that is $2.20 input, $0.55 cached and $6.60 output below 200k.
• Usage guideline violations: flagged requests are still charged; violations caught before generation in the Responses API carry a $0.05 fee per request.

Media, voice and tool pricing

For how input, output, cached and batch pricing work in general, see LLM Pricing Explained: Tokens and Caching. To cap spend on keys, see How to set spend limits on LLM API keys.

Where are xAI models available?

• First-party API: the xAI API, global endpoint and a US regional endpoint (grok-4.7 and grok-4.6 only, per the pricing page).
• Regions: the Grok 4.7 model page lists us-east-1; the Grok 4.20 reasoning page lists us-east-1 and us-west-2. The Grok 4.7 page says the model runs on multiple clusters.
• Grok 4.7 Fast: Cursor and Grok Build only, per the release notes.
• Open weights: none of the models on the current pricing page.
• Third-party clouds and consumer apps: not stated by xAI on the pages checked for this update.

Restricted-access xAI models

As of 2026-11-15. A named restricted-access programme for Grok models: not stated by xAI on the pages checked (2026-11-15). The table lists the access limits xAI does state.

What is the context window of xAI models?

Grok 4.7, Grok 4.6 and Grok 4.5 have 500k token context windows. Grok 4.3 and all Grok 4.20 variants have 1M. Grok Build 0.1 has 256k. For every text model, prompts of 200k tokens or more are billed at the long-context rate. See LLM Context Windows Explained.

Is any xAI model open source?

None of the Grok models on xAI's current pricing page are published as open weights; all are proprietary and accessed through the xAI API. See Open-Weight vs Closed LLMs Explained.

Model cards and system cards

xAI's model pages are specification pages (identifier, aliases, context, modalities, features, price, rate limits, regions). They do not include safety evaluation sections on the pages checked. See Model Cards and System Cards Explained.

Benchmarks

As of 15 November 2026. Provider-reported and independent results are kept in separate columns.

See How to Read LLM Benchmarks for how provider-reported and leaderboard figures differ.

Deprecations and retirements

As of 15 November 2026, from the release notes and models page.

Pinned dated identifiers stop working at retirement, while aliases such as grok-voice-latest move to the new model. Inventory where each identifier is used before the date. See the AI Model Retirement Guide.

Data handling and security

• Training use: xAI's security FAQ states that it never trains on API inputs or outputs without explicit permission, and does not train on the 30-day stored data.
• Default retention: API requests and responses are stored for 30 days for auditing in cases of suspected abuse or misuse, encrypted at rest, then deleted automatically.
• Zero data retention (ZDR): a team admin enables it in the xAI Console under Team Settings, after deleting existing Files and Collections; it applies team-wide, not per API key. With ZDR, prompts and outputs are not persisted to disk and the 30-day audit retention does not apply. The FAQ states that most customers should not enable it, because it disables features that depend on stored data, and points to [email protected] where the option is missing.
• Not available under ZDR: per-key request logging, the stateful Responses API, new uploads to the Files and Collections APIs, the Batch API, deferred completions and voice agent conversation history.
• Compliance: SOC 2 Type 2 is stated, with the Trust Center available to customers under NDA. A Business Associate Agreement can be requested through a BAA questionnaire.
• Key handling: generic controls apply to every provider's API: keep keys out of code, scope and rotate them, and set spend limits.

Source: xAI security FAQ (docs.x.ai/developers/faq/security), accessed 2026-11-15. See the LLMjacking Guide and the Foundation Model Selection Guide.

How to choose an xAI model: factual criteria

• Access: all listed Grok text models are on the public API except Grok 4.7 Fast; Grok Build 0.1 is early access.
• Retention requirements: ZDR disables the Batch API, the stateful Responses API and new Files and Collections uploads.
• Price per token: compare input, cached and output rates, and the 200k long-context step that applies to the whole request.
• Batch: 20% discount on listed text models; image and video generation billed at standard rates.
• Context: 500k on Grok 4.7, 4.6 and 4.5; 1M on Grok 4.3 and 4.20; 256k on Grok Build 0.1.
• Data residency: the US regional endpoint costs 1.1x and serves only grok-4.7 and grok-4.6.
• Rate limits: check the model page; Grok 4.7 lists 150 requests per second and Grok 4.20 reasoning lists 37.

Practitioner checklist

• Confirm the exact model ID and version on xAI's current models page before you build against it.
• Check the price per million input and output tokens on the official pricing page and the model card, and note the date you checked.
• Check the context window and maximum output for the specific model, not the model family.
• Read the deprecation and retirement notices and plan migration dates for any model you depend on.
• Review xAI's data handling, retention and training terms for the plan or platform you will use.
• Store provider API keys in a secrets manager and rotate them; never embed them in client code.
• Decide whether ZDR is needed before building on the Batch API or the stateful Responses API, which are unavailable under ZDR.
• Move workloads off grok-voice-2.0 to grok-voice-3.0 or grok-voice-latest.

Standards and references

• xAI: Models (accessed 2026-11-15)
• xAI: Grok 4.7 model page (accessed 2026-11-15)
• xAI: Grok 4.20 reasoning model page (accessed 2026-11-15)
• xAI: Release notes (accessed 2026-11-15)
• xAI: Pricing (accessed 2026-11-15)
• xAI: Security FAQ, docs.x.ai/developers/faq/security (accessed 2026-11-15)

Related NHI Mgmt Group resources: Mistral AI Models, LLM Security Guide, Foundation Model Selection Guide, LLMjacking Guide.