xAI Grok 4.20 Live: $2/$6 Pricing + 50% Batch Discount
xAI's April API update adds Grok 4.20, Grok 4.20 Multi-agent, Speech to Text, and broader Batch API support. Here's the pricing impact for developers.
By AI Pricing Guru Editorial Team
xAI’s developer docs now list Grok 4.20 and Grok 4.20 Multi-agent as live API models, alongside a broader Batch API update and newly available Speech to Text. The pricing headline is simple: Grok 4.20 costs $2.00 per million input tokens and $6.00 per million output tokens, with cached input at $0.20/M.
That puts xAI in an aggressive middle position: not as cheap as Grok 4.1 Fast, but far below the current flagship rates from OpenAI and Anthropic on output cost.
Grok 4.20 Pricing
| Model | Input ($/1M) | Cached input ($/1M) | Output ($/1M) | Notes |
|---|---|---|---|---|
| Grok 4.20 | $2.00 | $0.20 | $6.00 | Current general-purpose Grok 4 SKU |
| Grok 4.20 via Batch API | $1.00 | $0.10 | $3.00 | 50% token discount for async text jobs |
| Grok 4.1 Fast | $0.20 | $0.05 | $0.50 | Cheapest xAI text option |
| GPT-5.4 | $2.50 | $0.25 | $15.00 | OpenAI flagship comparison |
| Claude Opus 4.7 | $5.00 | $0.50 | $25.00 | Anthropic premium comparison |
For the full Grok rate card, see our xAI Grok pricing page. You can also model your own workload in the token cost calculator.
What changed in the xAI API?
The most important updates are:
- Grok 4.20 is live as the new default high-end Grok option in xAI’s model docs.
- Grok 4.20 Multi-agent is available for research-style tasks that coordinate multiple agents in real time.
- Batch API now supports image and video generation, not just chat completions.
- JSONL batch upload is available through the Files API, making large offline jobs easier to submit.
- Speech to Text is now generally available in 25 languages with batch and streaming modes.
- File and collection storage charges are active from April 20, 2026: $0.025/GiB/day for files and $0.10/GiB/day for collections.
For pricing buyers, the Batch API change is the one to pay attention to first. xAI says batch requests for text and language models are billed at 50% of standard token pricing and usually complete within 24 hours. That means a large offline Grok 4.20 workload can drop from $2/$6 to roughly $1 input / $3 output per 1M tokens.
One caveat: xAI says image and video generation can be submitted through the Batch API, but the 50% discount applies to text and language models only. Image and video jobs are supported operationally, not discounted.
What Multi-agent means for cost
Grok 4.20 Multi-agent is a different pricing risk than a regular chat model. xAI describes it as a beta feature that can coordinate multiple agents, use tools such as web search and X search, and synthesize a final answer through a leader agent.
The practical cost issue is that more agents means more reasoning and tool activity. xAI documents two setups:
- 4 agents for quicker, focused research
- 16 agents for deeper, more complex research
If you enable server-side tools, token pricing is only one part of the bill. xAI’s built-in tools add invocation charges — for example, web search, X search, and code execution are each $5 per 1,000 calls, while collections search is $2.50 per 1,000 calls.
So the safe budgeting rule is: use regular Grok 4.20 for normal chat, extraction, coding, and summarization; reserve Grok 4.20 Multi-agent for tasks where live research depth is worth extra latency and extra token/tool spend.
Who benefits?
High-volume async teams benefit most. If your workload can wait up to 24 hours — evaluations, data enrichment, content classification, offline summarization, synthetic data checks — Grok 4.20 via Batch API is now very competitive at the effective $1/$3 rate.
Research-agent builders also get a cleaner path. Grok’s biggest differentiation is live X and web grounding. Multi-agent makes that more useful for market monitoring, social analysis, competitive research, and news-heavy workflows.
Cost-sensitive real-time apps may still prefer Grok 4.1 Fast. At $0.20/$0.50, Grok 4.1 Fast remains the xAI model to test first for routing, extraction, and lightweight chat.
What to do now
If you’re already using xAI:
- Benchmark Grok 4.20 against your current Grok 4 or Grok 4.1 Fast prompts. The $6/M output rate is low enough to consider for production routing if quality holds.
- Move non-urgent text jobs to Batch API. A 50% discount is meaningful, especially for output-heavy summarization.
- Put caps around Multi-agent. Start with the 4-agent setup, track reasoning tokens and tool calls, and only move to 16 agents for tasks with clear business value.
- Audit file and collection storage. Storage charges are small, but persistent uploaded corpora can become a hidden line item.
Bottom line
xAI’s April update is not just a new model note. Grok 4.20 at $2/$6 — or effectively $1/$3 through Batch API for async text — gives developers a cheaper frontier-adjacent option with live-search advantages.
The tradeoff is predictability. Multi-agent and server-side tools can create variable bills if left uncapped. Treat Grok 4.20 as the new general-purpose xAI candidate, use Batch API wherever latency allows, and keep Grok 4.1 Fast in the routing stack for low-cost volume.
Related reading: compare the full market in our AI API pricing comparison, or see how the latest OpenAI launch changes the premium tier in GPT-5.5 pricing.