xAI Grok (4.6 / 4.3 / Grok Build): good value, but assess content safety before any student use

Short answer: Grok 4.6 costs about a fifth of the OpenAI flagship and has more than enough context and reasoning for school work, but inside Edor.ai it is reachable only through Poe, and its close tie to X platform content makes it the one American provider a school should formally assess before student use.

What this is: the newest of the four American providers

xAI was founded by Elon Musk in 2023 and is the youngest of the four major American providers. Grok 4, in July 2025, was the version that first put it into frontier-model contention, and it has iterated roughly quarterly since, holding its place in the conversation through 2026.

Two things set it apart from the other three. The first is pricing: at comparable capability, xAI's list rates sit consistently below OpenAI's and Anthropic's, with $2 / $6 working out at about a fifth of GPT-6 Astra. The second is positioning: Grok is deeply integrated with the X platform, X search is offered as a first-party tool, and the response style has always been presented as less restricted.

That second point is what a school has to deal with first. For a developer, fewer restrictions means more flexibility. For a school that answers to parents, it means an extra step of assessment. This is not a reason to rule Grok out; it is a reason to get the order right — assess, then enable, rather than enable and then observe.

The current line-up: three general tiers and two specialists

Grok 4.6 is the current flagship, released on 12 August 2026 with a 500K-token context window at $2 per 1M input tokens and $6 per 1M output tokens. It adds an xhigh reasoning level above low, medium and high, so long-chain tasks can ask the model to think for longer. There is a cost trap here identical to OpenAI's: once a request reaches 200K tokens, the whole request is billed at $4 / $12. The documentation also records Grok 4.6's knowledge cutoff as 1 February 2026, so without search tools enabled it knows nothing that happened afterwards.

Grok 4.5 (July 2026) is the direct predecessor, also 500K context at $2 / $6. Grok 4.3 (May 2026) is the longest-context and cheapest tier at 1M tokens and $1.25 / $2.50, which makes it the better choice for straightforward long-document work.

Two specialists complete the picture. grok-build-0.1 (June 2026) is the default model of the Grok Build coding agent, with a 256K context window at $1 / $2, which is worth knowing about when teaching programming in ICT. grok-voice-think-fast-2.0 (August 2026) is a speech-to-speech model, though as explained below the platform cannot currently reach it. The full list of versions, dates and prices is in the automatically generated series table further down this page.

Strengths: the four that matter most to a school

1. Clearly lower cost at comparable capability. At $2 / $6, high-volume marking and question generation become affordable without dropping to a small model that is not up to the job.
2. Useful context tiering. Use 4.6's 500K window when reasoning quality matters most, and switch to 4.3's cheaper 1M window when the task is simply to hold a lot of material. That division works better than always reaching for the flagship.
3. Reasoning adjustable up to xhigh. When marking answers that need point-by-point argument, as in Citizenship and Social Development or History, trading time for fuller reasoning is a genuinely usable control.
4. Transparent documentation. xAI's pricing page itemises the long-context doubling rule, server-side tool charges and batch discounts, which gives an IT coordinator a clear basis for a cost estimate.

Limits: what to say before anyone signs anything

• Content safety needs a school-level assessment. This is the most important point. There is a gap between Grok's response style and its X integration on one side, and what Hong Kong schools expect around minors on the other. The assessment should be completed and documented before use rather than retrofitted.
• It is not a native provider here. Grok is reachable only through Poe, which puts an additional third party in the path, means a privacy assessment must cover Poe as well, and leaves the school unable to manage xAI's account terms directly.
• Above 200K tokens the whole request doubles. A 500K window sounds generous, but crossing 200K steps the cost up a tier. Pasting a whole term of material in at once is poor value; retrieval over the relevant passages is the right approach.
• The voice model is out of reach. xAI publishes a speech-to-speech model, but Poe serves chat models only, so oral practice still depends on OpenAI or Azure OpenAI. The same applies to the embedding model a knowledge base needs.
• Little published evidence for Chinese classroom work. Compared with OpenAI and Claude, there is far less public testing of Grok on Hong Kong curriculum terminology and Traditional Chinese writing feedback. A school considering it for Chinese-language subjects should run a small comparison of its own first.

Where each tier fits

Teachers: bulk question generation, English writing feedback, marking that needs sustained argument such as data-response questions in Citizenship and Social Development, and comparing long documents. The low unit price pays off most on tasks that are high in volume.

Students: our advice is not yet. Until a content-safety assessment is complete and school management has agreed in writing, student conversations should stay on a provider that has already been assessed. If a school does proceed afterwards, keep the four-layer gating and teacher monitoring in place and set the tutor-mode hint ceiling more conservatively than the default.

IT coordinators: do not build the cost estimate on $2 / $6 alone. What actually has to be budgeted is the Poe plan, the doubling above the 200K threshold, and separate charges for server-side tools such as web search. Write the content-safety assessment up as a filed document as well; it is the thing you will want when a parent asks or an external review comes round.

Prompts worth testing

Copy these into a Edor.ai teacher tool or any Grok interface. The constraint lines are what make the output usable:

You are a Hong Kong secondary Citizenship and Social Development teacher. Below is a Secondary 4 student's data-response answer on population ageing in Hong Kong.
Output: first identify which claims are supported by the source material and which are bare assertions; then score Understanding of Sources, Argument and Expression out of 5 each with two comments per criterion; finish with three directions for improvement.
Constraints: write in Hong Kong English conventions; do not supply arguments on the student's behalf; do not introduce statistics beyond the source material provided; if you want to cite any figure whose origin you are unsure of, write "teacher to verify".
[paste sources and student answer]

You are a Hong Kong primary English teacher. Design a 20-minute reading comprehension worksheet for Primary 4 on the theme "a day in the life of a community cleaner".
Output: a passage of about 180 words, five comprehension questions (two literal retrieval, two inferential, one personal response), and a teacher marking guide.
Constraints: use the English conventions common in Hong Kong schools; keep the vocabulary at Primary 4 level; avoid proper nouns pupils would need a dictionary for; if you are unsure of the official English name of a Hong Kong occupation or department, write "teacher to verify" rather than inventing one.

Availability inside Edor.ai

xAI is not a native provider in Edor.ai. The five native routes are OpenAI, Azure OpenAI, Anthropic, Poe and local Ollama, so the only way to use Grok inside the platform is through the Poe provider. Once an administrator adds a Poe key, the Grok family can be selected from the supported model list; there is no separate xAI key field.

Before enabling it, we suggest completing a content-safety assessment signed off by the IT coordinator and the relevant panel heads. Grok's integration with X platform content is its most substantive difference from the other providers, and the gap between that and what Hong Kong schools expect around minors is a judgement the school has to make and record for itself. The platform cannot make it on the school's behalf. Once the assessment is done, a sound sequence is to limit Grok to teacher tools, gather a term of real observations, and only then discuss extending it.

Everything else matches the other providers reached through Poe. Calls still pass through the platform's four-layer gating and are recorded in ai_audit_logs, and because Poe carries no embedding or speech models, the school knowledge base and oral practice need a separate OpenAI or Azure OpenAI key.

Alternatives

• For a native provider whose single key covers chat, embedding and speech, see OpenAI.
• For a native provider with steadier writing feedback, see Anthropic Claude.
• For a lower unit price and fuller multimodality on the same Poe route, see Google Gemini.
• To understand how the four differ, read the four-provider comparison or hallucination and prompt-injection risk.