Grok 4.6: Release Date, Specs, Pricing, Benchmarks and Availability

The 500,000-token context window is useful for large codebases, long research material and multi-step agent sessions. It does not mean every request costs the same, however: xAI applies higher rates when a prompt reaches the long-context pricing band.

Grok 4.6 pricing

For API requests with prompts below 200,000 tokens, xAI lists these prices per 1 million tokens:

• Input: $2
• Cached input: $0.50
• Output: $6

For prompts of 200,000 tokens or more, the rates rise to $4 per million input tokens, $1 per million cached input tokens and $12 per million output tokens. xAI’s documentation says the higher rate applies to all tokens in a request that reaches that threshold.

That makes the short-context price headline attractive, but developers working with very large prompts should budget against the long-context rates rather than the $2/$6 figures alone.

How its benchmark score compares

Artificial Analysis gave Grok 4.6 a 61 on its Intelligence Index, a composite evaluation that places it level with GPT-5.6 Sol Max and five points above Grok 4.5 High.

That is a strong frontier position, but not an outright benchmark win. The same comparison places Grok 4.6 behind higher-scoring models on the overall index, while other evaluations show it trading wins and losses across coding, reasoning and knowledge-work tasks.

The more distinctive claim is efficiency. Grok 4.6’s standard API price is $2 per million input tokens and $6 per million output tokens, compared with $5 and $30 for GPT-5.6 Sol standard in the cited comparison. The evidence supports describing Grok 4.6 as a price-to-performance contender, not as the universally best model for every task.

Where Grok 4.6 is available

The model launched through the xAI API and xAI’s own developer products, including Grok Build and Cursor. xAI also listed access through partners such as OpenRouter, Vercel and Cloudflare.

Availability subsequently expanded to additional developer platforms:

• GitHub Copilot: Grok 4.6 rolled out for agentic coding and complex, multi-step workflows across supported Copilot plans and development environments.
• Amazon Bedrock: xAI announced general availability through Bedrock, with the 500,000-token context window and configurable reasoning settings.
• Grok’s consumer products: Grok remains available on grok.com and in the iOS and Android apps, although consumer access and usage limits depend on the applicable plan.

This distribution pattern reinforces the developer-oriented launch strategy: Grok 4.6 is being placed inside coding tools, cloud infrastructure and agent platforms, not only offered as a standalone chatbot.

What happened with the Grok Lite gibberish glitch?

Around August 20, some users of Grok Lite on Grok.com reported receiving long, nonsensical “word salad” responses instead of answers. Reports described the issue as affecting direct web queries, while the Grok experience on X appeared unaffected. Some users also said refreshing restored normal responses, although it did not work consistently for everyone.

Grok’s official account described the problem as a “rare temporary generation glitch,” said the service status page showed no broad incident, and advised affected users to start a fresh chat or regenerate the response.

The incident was presented as a temporary generation problem rather than evidence that Grok 4.6 itself was broadly unavailable. Still, it is a useful reminder to verify unusual model output—especially when a response suddenly becomes incoherent—and to retry in a new session before relying on it.

What xAI’s positioning suggests

Taken together, the launch details point to a clear competitive strategy:

1. Target agent builders. The emphasis on Grok Build, coding integrations, tool use and long-running tasks is aimed at developers building software agents.
2. Compete on economics. A 61 Intelligence Index score paired with lower standard API rates gives xAI a price-to-performance story, even where the model does not lead every benchmark.
3. Use distribution to reduce friction. Access through APIs, coding environments, Copilot, Bedrock and other partners makes Grok 4.6 easier to test inside existing workflows.
4. Sell practical capability rather than only a headline score. The 500,000-token context window and configurable reasoning are particularly relevant to repository-scale coding, research and multi-step automation.

The cautious conclusion is that Grok 4.6 is positioned as an economical frontier model for production-oriented agents and developer workflows. That is an inference from xAI’s product design, pricing and distribution—not proof that it will outperform every rival in every real-world task.