Grok V9-Medium: What xAI's 1.5T Coding Model Means for Free Users

On June 10, 2026, xAI published release notes for Grok V9-Medium, a 1.5 trillion parameter coding model. The company put the model in the free tier immediately, but it cut free coding requests at the same time. xAI said the model uses a mixture-of-experts design with 420 billion active parameters and a 256,000 token context window. The release notes appeared on Hugging Face alongside the model weights. That made the launch both a free-tier expansion and a rate-limit squeeze. Free users gained access to a much stronger model, but heavy coders lost capacity. The change follows a broader shift documented in Free AI News.

The free tier now includes 15 coding requests per four-hour window for Grok V9-Medium. That is down from 30 requests on the previous Grok 4.1 coding mode. Paid SuperGrok users, at $30 per month, receive 300 requests in the same window. Free output caps are 8,192 tokens per response, while paid users get 16,384 tokens. Free users can still send general chat messages, but coding sessions count separately. xAI set the limits in its June 10 pricing page. The change hits solo developers who used Grok 4.1’s free coding mode for daily work. It also pushes teams toward the paid plan. Our breakdown of xAI subscription tiers covers the full pricing shift.

Grok V9-Medium arrived during a heated coding-model pricing fight. OpenAI had already expanded Codex free tier access, while Anthropic moved Claude Code to credit pools in June. xAI’s move undercut both on model size and free access, but not on usable volume. The free tier is generous enough for a few tests, not for ongoing development. Independent data from Stanford HAI shows coding is the fastest-growing enterprise AI use case. That makes free coding limits a strategic lever. xAI wants users to try flagship-level code generation, then pay when they hit the wall. Free AI News reported on the OpenAI Codex free tier and Anthropic’s credit overhaul.

June 10 also marked a quieter change. xAI stopped offering Grok 4.1 coding mode to new free accounts. Existing free users kept access until June 17, 2026, then the old mode was removed. That meant the only free coding path was Grok V9-Medium with tighter limits. The pattern matched GitHub Copilot’s usage-based billing push and Google’s Gemini quota changes. Users now face a choice: accept 15 requests per four hours or pay. The release signals that free AI coding is becoming a sampler, not a daily tool. For a wider view, see our June free tier coverage.

How Do the Top Options Compare?

Limits reflect xAI’s June 10, 2026 release notes and competitor free tier pages as of June 12, 2026. Free AI News may earn a commission if you sign up for a paid plan through links on this site. This does not affect our reporting.

1. Grok V9-Medium Free Tier , Occasional free code generation

xAI launched Grok V9-Medium on June 10, 2026 and put it in the free tier the same day. The model has 1.5 trillion total parameters, with 420 billion active under a mixture-of-experts design. The free context window is 256,000 tokens and the output cap is 8,192 tokens. Users can send 15 coding requests per four-hour window. That is down from 30 on the previous Grok 4.1 coding mode, which xAI removed for new free accounts on June 10.

The free tier gives users access to a model that xAI claimed beat Grok 4.1 on HumanEval and SWE-bench by at least 20%. But the request limit is tight. A single multi-step debugging session can consume five requests in minutes. Once the limit is hit, users must wait up to four hours. xAI’s model card on Hugging Face lists the weights as open under a modified attribution license, and a mirror appeared on GitHub. Running the full 1.5T model locally requires hardware most free users do not have.

For occasional code generation, the free tier is a clear upgrade. For daily coding, it is not a replacement. The same tension appears across major AI free tiers. Free users should treat Grok V9-Medium as a trial, not a workhorse.

Key strengths:

• ✅ Access to a 1.5 trillion parameter coding model without payment
• ✅ 256,000 token context window covers large code files
• ✅ Weights posted on Hugging Face for local use
• ✅ Better benchmark scores than previous Grok coding mode
• ❌ Only 15 coding requests per four hours
• ❌ Output cap of 8,192 tokens can truncate long code responses
• ❌ Grok 4.1 coding mode removed for new free users

Who it’s for: Developers who need occasional code snippets or want to test a large open model without paying.

2. Grok SuperGrok Paid Plan , Daily coding and larger agent loops

SuperGrok costs $30 per month and includes 300 coding requests per four-hour window for Grok V9-Medium. That is a 20x increase over the free tier. Paid users also get priority inference during peak hours and a 16,384 token output cap. The plan launched alongside the free tier on June 10, 2026.

The paid plan is aimed at developers who run test suites, refactor large files, or use coding agents. At 300 requests per four hours, a solo developer can maintain a steady workflow without hitting the reset wall. xAI also kept general chat unlimited on SuperGrok, while coding requests are metered. The pricing details appeared in xAI’s June 10 pricing page, which Free AI News covered in its subscription comparison.

SuperGrok undercuts GitHub Copilot Pro and matches ChatGPT Plus on price. But it lacks the full IDE integration that GitHub Copilot and Cursor offer. Users who want agentic coding with xAI can connect via the Grok Build API. That API has its own per-token fees, separate from the SuperGrok plan.

Key strengths:

• ✅ 20x more coding requests than the free tier
• ✅ Priority access reduces queue wait during peak hours
• ✅ 16,384 token output cap allows longer completions
• ✅ Competitive $30 monthly price
• ❌ Coding requests still reset on a four-hour window
• ❌ No native first-party IDE plugin at launch
• ❌ Agentic API use is billed separately from the plan

Who it’s for: Developers who need Grok V9-Medium for daily coding work without per-token API billing.

3. OpenAI Codex Free Tier , ChatGPT users who want free code help with occasional limits

OpenAI expanded Codex free tier access in May 2026, offering 10 coding requests per five-hour window. The free tier runs on a smaller Codex model with a 128,000 token context window and 4,096 token output limit. Free users access it inside ChatGPT, not through a standalone coding IDE.

Compared with Grok V9-Medium, the Codex free tier has fewer requests and a smaller context window. But OpenAI’s model is simpler to use inside ChatGPT and supports memory features for free users. xAI’s 15 requests per four hours is more generous on volume, but OpenAI’s five-hour reset window can be easier to plan around. Free AI News covered the OpenAI Codex free tier expansion.

OpenAI also maintains a separate Codex agentic coding product for paid users. That split keeps free tier coding constrained, much like xAI’s approach. For users deciding between the two, the AI coding pricing guide compares both.

Key strengths:

• ✅ Integrated into ChatGPT, no separate app needed
• ✅ Five-hour reset window gives predictable scheduling
• ✅ Free tier includes memory features for coding context
• ❌ Only 10 coding requests per five-hour window
• ❌ 128,000 token context window is half of Grok’s
• ❌ Output cap of 4,096 tokens is low for long code files

Who it’s for: Users already inside ChatGPT who need occasional code help and prefer OpenAI’s tooling.

4. Anthropic Claude Code Free Tier , Users who need careful code reasoning but can accept tight limits

Anthropic moved Claude Code to a credit pool model on June 15, 2026. Free users receive a small set of console credits per month, roughly equivalent to 5 coding requests per five-hour window. The free tier runs on Claude Opus 4.8 in fast mode, with a 200,000 token context window and 4,000 token output limit.

Anthropic’s free tier is the most restrictive among major coding assistants. The credit pool replaced the flat-rate access that free users previously had. That change drew anger from developers and was covered in our report on Anthropic’s credit overhaul. The Claude free tier also resets on a five-hour basis, but the monthly credit cap makes heavy use impossible.

Grok V9-Medium offers three times the free request volume and a larger context window. However, Claude Code’s reasoning quality remains strong for complex debugging. Users who value careful analysis over volume may still prefer Anthropic, even with the tighter limits.

Key strengths:

• ✅ Claude Opus 4.8 fast mode is strong on code reasoning
• ✅ 200,000 token context window is solid
• ✅ Free credits can be saved and used in bursts
• ❌ Only about 5 coding requests per five-hour window
• ❌ Monthly credit cap prevents sustained work
• ❌ Output limit of 4,000 tokens truncates longer code

Who it’s for: Developers who need high-quality reasoning for occasional complex problems and can accept severe limits.

5. Google Gemini Code Assist Free Tier , Developers who want the highest free request volume

Google cut Gemini Code Assist’s free tier in June 2026 but left 25 coding requests per four-hour window. The free tier runs on Gemini 2.0 Flash, not the larger Pro model. Context window is 1 million tokens, and output cap is 8,000 tokens. That makes Google the volume leader on free coding requests.

Google’s approach differs from xAI. Free users get more requests but a smaller model. Grok V9-Medium is larger and stronger on code benchmarks, but xAI gives fewer requests. Google also tightens compute quotas for agentic use, which triggered a backlash covered in Free AI News. The free tier works inside Google AI Studio and Colab.

For users who need many small code snippets, Gemini Code Assist free is the best option. For users who need a frontier-size model for complex refactoring, Grok V9-Medium’s free tier is stronger. Both signal that free coding is now a funnel to paid plans. Free AI News may earn a commission if you sign up for a paid plan through links on this site. This does not affect our reporting.

Key strengths:

• ✅ 25 coding requests per four-hour window, highest among major free tiers
• ✅ 1 million token context window handles entire repositories
• ✅ No credit card required for Google AI Studio access
• ❌ Runs on Gemini 2.0 Flash, not the stronger Pro model
• ❌ Agentic compute quotas can lock free users out
• ❌ Output cap of 8,000 tokens can cut long code files

Who it’s for: Developers who need frequent free code help and work inside Google AI Studio or Colab.

Frequently Asked Questions

What Should You Remember?

• Free tier: xAI added Grok V9-Medium to the free tier but cut coding requests from 30 to 15 per four-hour window.
• Pricing: SuperGrok costs $30 per month and includes 300 coding requests per four-hour window.
• Model: Grok V9-Medium has 1.5 trillion total parameters and 420 billion active parameters.
• Context: Free users get a 256,000 token context window and an 8,192 token output cap.
• Competition: Grok’s free tier offers more requests than OpenAI Codex but fewer than Google’s 25 requests per four hours.
• Old tier removed: xAI removed Grok 4.1 coding mode for new free accounts on June 10, 2026.
• Takeaway: Free coding tiers are now samplers, not daily tools.

Free AI News is an independent editorial publication. Information about AI pricing, free-tier limits, and features changes frequently and may become outdated. Always verify current details through the vendor’s official pages. Affiliate links may earn a commission at no cost to you, and never affect our reporting.