Grok 4.6 gives you frontier-level intelligence at $2/$6 per 1M tokens — a fifth of GPT-5.6 Sol's output price — plus live X data nothing else can match. The safety record is the catch.
Grok 4.6 (August 12, 2026) is SpaceXAI's best model yet — it scores 61 on the Artificial Analysis Intelligence Index, level with GPT-5.6 Sol and close to Claude Opus 5, while the API costs just ~$2 per 1M input and $6 per 1M output tokens (GPT-5.6 Sol is $5/$30). There is a real free tier on grok.com, and paid plans run from ~$10/month (SuperGrok Lite) to ~$30 (SuperGrok) and ~$300 (Heavy). Buy it for live X and web search, agentic cost efficiency and personality. Don't rely on it alone for regulated or safety-sensitive work: 2026 brought sustained criticism of Grok's guardrails and generated imagery, it trails on conventional coding benchmarks (DeepSWE 65.9% vs 73% for GPT-5.6 Sol), and its knowledge cutoff is February 1, 2026. Free alternatives: Gemini, DeepSeek and Perplexity.
Grok is the AI assistant built by xAI — the company Elon Musk founded in 2023, which SpaceX acquired in February 2026 at a reported $250 billion valuation and rebranded as SpaceXAI in July 2026. The product line kept the Grok name. It is available at grok.com, inside the X app, as iOS and Android apps, as the Grok Build coding surface, inside Cursor, and through a pay-per-token developer API.
The current flagship is Grok 4.6, released on August 12, 2026, one month after Grok 4.5 (July 8, 2026). SpaceXAI describes it as focused on long-running agents and more ambitious interactive and visual work — that is, keeping a task coherent across many steps rather than winning one-shot trivia. The model ships with a 500k token context window and a training knowledge cutoff of February 1, 2026.
What makes Grok structurally different from ChatGPT, Claude and Gemini is not the model weights — it is the data pipe. Grok has native, server-side access to X (Twitter) search alongside web search, so it can see breaking posts, replies and sentiment that competitors cannot index the same way. That single feature is why a lot of people keep Grok installed even when they pay for something else.
Verified from x.ai's official pricing and model pages, Artificial Analysis and pricing trackers (September 2026). Grok is sold across seven consumer tiers plus a separate prepaid API:
Free — ~$0: grok.com and apps, lighter Grok model, limited chat + voice allowance, basic real-time web and X search. No image or video generation.
SuperGrok Lite — ~$10/month: basic Grok Imagine (480p, ~6-second clips), about 2x the free chat allowance, one AI agent.
SuperGrok — ~$30/month (or ~$300/year, about $25/month effective): Grok 4.6, Grok Bot, connectors, higher limits, image and video generation.
X Premium — ~$8/month / X Premium+ — ~$40/month: X subscriptions with Grok bundled at varying limits. Only worth it if you already want X.
SuperGrok Plus — ~$100/month: 1080p video, significantly higher usage across Chat, Imagine, Voice and Build, priority access, early features.
SuperGrok Heavy — ~$300/month: largest usage pool, Grok 4 Heavy, multi-agent work, Grok Bot.
API — Grok 4.6: ~$2 / 1M input tokens, ~$6 / 1M output tokens (cached input ~$0.50). Prompts at or above 200k tokens bill at ~$4/$12. Grok 4.3 stays cheaper at ~$1.25/$2.50 with a 1M context.
Prices are USD and change quickly — some tiers are regional and taxes vary. All paid consumer plans share one weekly usage pool (changed in June 2026) instead of per-product daily caps. Confirm current pricing on x.ai before subscribing.
The headline number to remember is the API's $2/$6 per 1M tokens. Independent testing measured Grok 4.6 at about $0.84 per task on the Intelligence Index — the same cost as Kimi K3 but with more intelligence — which places it on the cost-versus-performance Pareto frontier against Claude Opus 5 at $5/$25 and GPT-5.6 Sol at $5/$30. It is also unusually token-efficient on long work: roughly 53 turns and 0.5B input tokens to finish tasks that took Claude Opus 5 around 103 turns and 2.0B input tokens, and about a third of terminal-based tasks cost the same tokens as a strong rival.
Yes — and unlike OpenAI's flagship experience, the free tier is not a demo. On grok.com and the mobile apps you get a lighter Grok model, a limited chat and voice allowance, and basic real-time web and X search, at no cost and without jumping through hoops. Two limits matter: image and video generation are no longer included free (they were pulled from the free tier), and free allowances reset on their own cycle rather than drawing on the paid weekly pool.
If you want paid Grok without paying for X, SuperGrok Lite at around $10/month is the cheapest entry point — cheaper than ChatGPT Plus ($20) and well below the $40 X Premium+ bundle. That makes Grok the value pick of the big four assistants at the low end, provided you are comfortable with its safety record (more on that below).
Grok 4.6's launch numbers come largely from the vendor, but Artificial Analysis independently reproduced the headline result. The pattern is consistent: strong and efficient on agentic work, mid-pack on classic reasoning and coding benchmarks.
Read plainly: Grok 4.6 is a near-frontier model at a mid-market price, not a new leader. If your work is "research this, then build a first version of that", it is an excellent buy. If your work is hard, long-form software engineering, GPT-5.6 and Claude still score higher — and you will feel it.
If Grok's safety record, coding benchmarks or ecosystem is a blocker, these are the realistic swaps — including the free ones.
Best for: the safest all-round choice — our full ChatGPT review
Best for: the most generous free tier — our full Gemini review
Best for: writing quality and long-horizon agentic work — our full Claude review
Best for: 100% free reasoning — our full DeepSeek review
Best for: cited research where you need to verify — our full Perplexity review
| Tool | Free tier | Cheapest paid | API (in / out per 1M) | Best for |
|---|---|---|---|---|
| Grok (4.6) | Yes — limited chat/voice, no image gen | ~$10/mo (Lite) | ~$2 / $6 | Live X data, agentic cost efficiency |
| ChatGPT (GPT-5.6) | Limited, ad-supported | ~$20/mo (Plus) | ~$5 / $30 | All-round use and ecosystem |
| Claude (Opus 5) | Limited free tier | ~$20/mo (Pro) | ~$5 / $25 | Writing and long-horizon work |
| Gemini | Yes — effectively unlimited | ~$20/mo (AI Pro) | Varies by tier | Free assistance + Google apps |
| DeepSeek | Yes — unlimited | Free | Very cheap | Free reasoning, maths and code |
| Perplexity | Yes, limited | ~$20/mo (Pro) | N/A (consumer product) | Cited, verifiable research |
Free at $0; SuperGrok Lite around $10/month; SuperGrok around $30/month (about $300/year); SuperGrok Plus around $100/month; SuperGrok Heavy around $300/month. X Premium (~$8) and X Premium+ (~$40) bundle Grok with X. The Grok 4.6 API is about $2 per 1M input and $6 per 1M output tokens, with cached input around $0.50. Prices change often — confirm on x.ai.
Yes, there is a real free tier on grok.com and the mobile apps: a lighter model, a limited chat and voice allowance and basic real-time web and X search. Image and video generation through Grok Imagine are no longer included free, and free allowances reset on their own cycle rather than the paid weekly pool.
Grok 4.6, released August 12, 2026 — one month after Grok 4.5. It focuses on long-running agents, coding and visual/interactive work, has a 500k token context window and a training cutoff of February 1, 2026, so real-time answers require web or X search to be enabled.
Not overall — closeness at a much lower price is the real story. Independent testing scores Grok 4.6 at 61 on the AA Intelligence Index, level with GPT-5.6 Sol (61) and behind Claude Opus 5 (63), while costing roughly $2/$6 per 1M tokens against $5/$30 for GPT-5.6 Sol. Grok is strong agentically and very token-efficient, but trails on conventional coding (DeepSWE 65.9% vs 73%) and newer terminal benchmarks.
Real-time information, cost efficiency and personality. Its live X search is unique, it is strong at taking a rough idea to a working first version, and at $2/$6 per 1M tokens it sits on the cost-versus-performance frontier for agentic work — measured at about $0.84 per task, similar to Kimi K3 with higher intelligence.
Gemini (most generous free tier), DeepSeek (free, strong reasoning and maths), ChatGPT and Claude (limited free tiers) and Perplexity (free cited research). If live social data is the reason you want Grok, nothing else matches it — but for general free use, Gemini and DeepSeek give you more.
Honest accounting software reviews: Wave, QuickBooks, Xero & free options.
🎮 Gaming Gear ReviewsEsports hardware decoded: gaming mice, keyboards, headsets & budget picks.
📄 PDF Merge OnlineCombine multiple PDF files into one — free, no upload, files stay on your device.
📷 ID Photo StudioPassport & visa photos with AI background changer, compressor & converter.
🧮 HK Salary Tax CalculatorEstimate Hong Kong salaries tax for 2026/27 in seconds.
🎁 HK Benefits Quick CheckGovernment & benefits eligibility checker — OAA, CSSA, housing & more.
📅 HK Calendar 香港月曆Public holidays, lunar dates & solar terms for 2020-2036 — bilingual, free, no sign-up.