On May 6, 2026, SpaceXAI confirmed that its Grok model family landed on AWS Bedrock. Developers could call Grok 4, Grok 4.1 Fast, and Grok Code Fast directly through Bedrock APIs. The move put SpaceXAI’s models inside Amazon’s managed AI service for the first time. Pricing started at $0.25 per million input tokens for the fast model. A limited free tier ran through June 30, 2026, with caps on daily requests. Full pricing comparison shows where these rates sit.
The change hit enterprise developers who already used Bedrock for other frontier models. Those teams no longer needed a separate xAI API key or a direct billing relationship with SpaceXAI. Instead, they could consolidate spend inside AWS and use existing IAM policies. Smaller builders also got a new path to Grok models without a paid xAI subscription. But the free tier was not unlimited. AWS capped each account at 100 requests per day for the fast model and 20 requests per day for the full Grok 4. Anthropic has pushed a similar consolidation angle for its Claude models on cloud marketplaces.
Why this matters is obvious to anyone tracking model pricing. SpaceXAI undercut current Bedrock rates for comparable reasoning models. Google’s Gemini Pro API charged about $1.25 per million input tokens at the time. OpenAI’s GPT-5 API on Azure was near $2.00. Bedrock’s Grok 4 pricing at $0.75 per million input tokens reset the floor. Google AI had been cutting prices, but SpaceXAI moved faster. This forced existing providers to respond.
The announcement came from SpaceXAI’s official blog and AWS Bedrock’s model catalog page. Both confirmed general availability as of May 6, 2026. US East, US West, and EU regions got access first. Asia Pacific rollout followed on May 20, 2026. The free tier applied to all new Bedrock accounts. Existing AWS customers could opt in immediately. AI API free tier limits have tightened across major providers. This launch tested how far a free tier could go.
How Do the Top Options Compare?
| Offering | Best For | Input Price | Output Price | Free Tier |
|---|---|---|---|---|
| Grok 4 on Bedrock | High-complexity reasoning and agentic coding | $0.75 per 1M tokens | $3.00 per 1M tokens | 20 requests/day until June 30 |
| Grok 4.1 Fast on Bedrock | Low-latency chat and classification | $0.25 per 1M tokens | $1.00 per 1M tokens | 100 requests/day until June 30 |
| Grok Code Fast on Bedrock | Code generation and refactoring | $0.40 per 1M tokens | $1.60 per 1M tokens | 50 requests/day until June 30 |
| xAI API Direct | High rate limits and early access | $0.70 per 1M tokens | $2.80 per 1M tokens | No free tier |
Prices listed are per million tokens. Free tier limits reset daily at midnight UTC. Bedrock customers pay additional AWS infrastructure charges for provisioned throughput.
1. Grok 4 on AWS Bedrock , High-complexity reasoning, long-context analysis, and agent workflows
SpaceXAI’s full Grok 4 model came to Bedrock with a 256k context window and tool-use support. Developers could call it through the Bedrock Converse API or InvokeModel. The model handled multi-step reasoning, code generation, and structured JSON output. AWS listed it under Frontier Models in the Bedrock catalog. Full subscription tier comparison puts Grok 4 against OpenAI and Anthropic plans.
Pricing landed at $0.75 per million input tokens and $3.00 per million output tokens. That price undercut OpenAI’s GPT-5 on Azure by roughly 60 percent. It also beat Google’s Gemini 2.5 Pro API rate at the time. The free tier allowed 20 requests per day until June 30, 2026. After that, accounts moved to standard pay-as-you-go rates. OpenAI did not match that free allowance.
Enterprise teams liked the IAM integration and existing VPC controls. Startups gained a cheaper path to a frontier reasoning model. But the daily free limit was tight. Many test workloads exhausted 20 requests in under an hour. Provisioned throughput pricing started at $120 per hour for dedicated capacity.
Key strengths:
- ✅ 256k context window supports long documents and codebases
- ✅ Tool-use support enables agentic workflows on Bedrock
- ✅ Input price undercuts comparable OpenAI and Google models
- ✅ Native AWS IAM and VPC policies reduce compliance burden
- ❌ Free tier capped at 20 requests per day, making testing slow
- ❌ Output tokens cost 4x the input rate
- ❌ Provisioned throughput has a high minimum hourly charge
Who it’s for: Teams that need a frontier reasoning model inside AWS with existing security controls should pick Grok 4 on Bedrock.
2. Grok 4.1 Fast on AWS Bedrock , Low-latency responses, high-volume classification, and interactive chat
Grok 4.1 Fast targeted latency-sensitive workloads. The model returned responses in under 400 milliseconds for short prompts on Bedrock’s default throughput. It supported a 64k context window, which covered most support tickets and chat logs. Developers used it for real-time moderation and customer service routing. Grok V9 Medium free users saw a related free tier expansion from SpaceXAI.
The price was the biggest story. At $0.25 per million input tokens and $1.00 per million output tokens, Grok 4.1 Fast became the cheapest frontier-class model on Bedrock. That rate beat Google’s Gemini Flash API on some tiers. It was less than half the price of Anthropic’s Claude Sonnet on Bedrock. Google AI had cut Gemini Flash prices earlier in 2026, but SpaceXAI set a new floor.
The free tier was much more usable at 100 requests per day. Small developers could build and test without paying until June 30, 2026. After that, AWS charged standard per-token rates. Rate limits rose for paid accounts to 500 requests per minute in US East.
Key strengths:
- ✅ Input price of $0.25 per 1M tokens is the lowest on Bedrock
- ✅ 100 daily free requests allow real testing
- ✅ Fast response times fit chat and routing use cases
- ✅ 64k context window handles most production documents
- ❌ Smaller context window than Grok 4
- ❌ Output token price may add up for long generation tasks
- ❌ Free tier expires June 30, 2026
Who it’s for: Developers who want cheap, fast Grok access for high-volume or latency-sensitive apps should choose Grok 4.1 Fast.
3. Grok Code Fast on AWS Bedrock , Code generation, refactoring, and agentic coding assistants
Grok Code Fast was the third model SpaceXAI placed on Bedrock. It specialized in code completion, bug fixes, and test generation. The model supported function calling and streaming, which made it easy to plug into CI/CD pipelines. AWS documented it under Code Generation in the Bedrock catalog. Grok Build 01 showed SpaceXAI’s earlier coding model push, but this Bedrock version added managed scaling.
Pricing came in at $0.40 per million input tokens and $1.60 per million output tokens. That was more expensive than Grok 4.1 Fast but cheaper than full Grok 4. The free tier allowed 50 requests per day through June 30, 2026. Teams could test the model on small repos without initial cost. After the free period, standard Bedrock rates applied. GitHub Copilot usage-based billing has pushed developers to compare code models by token cost.
One limitation was the 32k context window. Large monorepos needed chunking or retrieval. But the model’s speed was notable. Average code completion latency on Bedrock was under 300 milliseconds. AWS promised 99.9 percent availability for the model in US regions.
Key strengths:
- ✅ Fast code completion with under 300ms average latency
- ✅ Function calling and streaming fit CI/CD workflows
- ✅ Free tier of 50 requests per day allows repo testing
- ✅ Price sits between fast and full Grok models
- ❌ 32k context window is too small for large monorepos
- ❌ Output cost can rise quickly for long code generation
- ❌ No on-device or local deployment option
Who it’s for: Software teams that need managed code generation inside AWS should select Grok Code Fast on Bedrock.
4. xAI API Direct , High rate limits, custom fine-tuning, and early access to new Grok versions
SpaceXAI still sold direct API access outside AWS Bedrock. The direct service offered higher default rate limits and no AWS middleware overhead. Developers who needed dedicated capacity or custom fine-tuning kept using xAI’s own dashboard. Pricing was similar to Bedrock for Grok 4, at $0.70 per million input tokens and $2.80 per million output tokens. AI API free tier limits changed across providers in 2026, and xAI direct had no free tier.
Direct access gave teams first access to new Grok releases before Bedrock. That mattered for companies testing new model features. But the direct route required separate billing and lacked AWS IAM integration. Many enterprise buyers moved to Bedrock to consolidate compliance and procurement. Anthropic faced the same split between direct API and cloud marketplaces.
Rate limits were the main advantage. Paid direct customers received 2,000 requests per minute for Grok 4.1 Fast, four times the Bedrock standard. High-volume startups often paid the direct price for this headroom. Cost-sensitive teams preferred Bedrock’s free tier.
Key strengths:
- ✅ Higher rate limits than Bedrock standard tiers
- ✅ Earlier access to new Grok model versions
- ✅ No AWS infrastructure markup
- ✅ Direct fine-tuning and dedicated capacity options
- ❌ No free tier for testing
- ❌ Separate billing and compliance burden
- ❌ No native AWS IAM or VPC integration
Who it’s for: High-volume teams that need maximum throughput and direct SpaceXAI support should use xAI API Direct.
Frequently Asked Questions
When did SpaceXAI Grok models become available on AWS Bedrock?
SpaceXAI confirmed availability on May 6, 2026. AWS updated its Bedrock model catalog the same day. US East, US West, and EU regions got access first. Asia Pacific followed on May 20, 2026.
Which Grok models are on Bedrock?
Grok 4, Grok 4.1 Fast, and Grok Code Fast are available. Each has different context windows and pricing. Developers can call them through the Bedrock Converse API or InvokeModel.
What are the prices for Grok models on Bedrock?
Grok 4 costs $0.75 per million input tokens and $3.00 per million output tokens. Grok 4.1 Fast costs $0.25 input and $1.00 output. Grok Code Fast costs $0.40 input and $1.60 output. Prices exclude AWS infrastructure charges.
Is there a free tier?
Yes, a limited free tier ran through June 30, 2026. Grok 4 allowed 20 requests per day, Grok 4.1 Fast allowed 100 requests per day, and Grok Code Fast allowed 50 requests per day. After June 30, accounts moved to standard pay-as-you-go rates.
How does Bedrock pricing compare to direct xAI API?
Direct xAI API pricing is similar but has no free tier. Grok 4 direct costs $0.70 per million input tokens and $2.80 per million output tokens. Bedrock adds a small markup but includes AWS billing, IAM, and VPC controls.
Who should use Bedrock instead of direct xAI API?
Teams that already use AWS and need compliance, IAM, and consolidated billing should use Bedrock. High-volume teams that need 2,000 requests per minute or custom fine-tuning may prefer direct xAI API. Both offer the same core Grok models.
What Should You Remember?
- Availability: SpaceXAI placed Grok 4, Grok 4.1 Fast, and Grok Code Fast on AWS Bedrock on May 6, 2026.
- Pricing: Grok 4.1 Fast costs $0.25 per million input tokens, the lowest frontier-class rate on Bedrock at launch.
- Free tier: New Bedrock accounts got 20 to 100 daily requests free until June 30, 2026.
- Context windows: Grok 4 supports 256k tokens, Grok 4.1 Fast supports 64k, and Grok Code Fast supports 32k.
- Competitive pressure: SpaceXAI undercut Google Gemini and OpenAI API rates by roughly 60 percent on input tokens.
- Direct alternative: xAI direct API had no free tier but offered higher rate limits and earlier model access.
- Enterprise shift: AWS IAM and VPC integration pulled existing Bedrock customers away from separate xAI billing.
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.
