tokenmix · 2026-04-03
xAI's Grok API is one of the most aggressively priced in 2026. Grok 4.1 Fast costs $0.20/M input — matching DeepSeek V4 territory — while offering a 2-million-token context window that no other model at this price point can touch. The flagship Grok 4.20 sits at $2.00/$6.00, undercutting GPT-5.4 on output by 60%. New accounts get $25 in free credits, plus an optional
All prices per 1M tokens, xAI direct API, April 2026:
| Model | Input | Cached Input | Output | Context | Best For |
|---|---|---|---|---|---|
| Grok 4.1 Fast Reasoning | $0.20 | $0.05 | $0.50 | 2M | Budget production, agents |
| Grok 4.1 Fast Non-Reasoning | $0.20 | $0.05 | $0.50 | 2M | Simple tasks, high volume |
| Grok 4.20 Reasoning | $2.00 | $0.20 | $6.00 | 2M | Flagship quality |
| Grok 4.20 Non-Reasoning | $2.00 | $0.20 | $6.00 | 2M | Fast flagship, no CoT |
| Grok 4.20 Multi-Agent | $2.00 | $0.20 | $6.00 | 2M | Multi-agent orchestration |
Image & Video generation:
The headline: Every Grok model has a 2M token context window — the largest in the industry at any price point. GPT-5.4 offers 1.1M. Claude offers 1M. Grok doubles them both.
Grok 4.1 Fast is xAI's answer to DeepSeek V4 and GPT-5.4 Nano — budget pricing with flagship-tier context.
| Spec | Grok 4.1 Fast | DeepSeek V4 | GPT-5.4 Nano |
|---|---|---|---|
| Input/M | $0.20 | $0.30 | $0.20 |
| Output/M | $0.50 | $0.50 | .25 |
| Cached Input/M | $0.05 | $0.03 | $0.02 |
| Context Window | 2M | 1M | 400K |
| Reasoning mode | Yes | No | No |
Grok 4.1 Fast ties GPT Nano on input, beats it 2.5x on output, and has 5x the context window. It also offers a reasoning mode toggle — something neither DeepSeek V4 nor GPT Nano has at this price.
Against DeepSeek V4: Grok is 33% cheaper on input ($0.20 vs $0.30), identical on output ($0.50), and has double the context (2M vs 1M). DeepSeek's cache hit is cheaper ($0.03 vs $0.05), but for everything else Grok wins.
Best for: Agent workflows that need massive context at budget pricing. The 2M window handles entire codebases without chunking.
Grok 4.20 is xAI's premium model, released March 2026. The standout metric is output pricing.
| Spec | Grok 4.20 | GPT-5.4 | Claude Sonnet 4.6 | Gemini 3.1 Pro |
|---|---|---|---|---|
| Input/M | $2.00 | $2.50 | $3.00 | $2.00 |
| Output/M | $6.00 | 5.00 | 5.00 | 2.00 |
| Cached Input/M | $0.20 | $0.25 | $0.30 | $0.50 |
| Context | 2M | 1.1M | 1M | 1M |
Grok 4.20 output is 60% cheaper than GPT-5.4 and Claude Sonnet ($6 vs 5). For output-heavy workloads — content generation, code writing, detailed analysis — this is the single biggest price difference among flagship models.
Combined with Mistral Large 3 ($2/$6 output), Grok 4.20 proves that the 5/M output standard set by OpenAI and Anthropic is no longer the floor for frontier-quality models.
TokenMix.ai tracks these cross-provider pricing gaps in real time — check tokenmix.ai/pricing for the latest rates.
xAI offers two free credit programs:
| Program | Amount | Requirement |
|---|---|---|
| Signup bonus | $25 | New account, no credit card needed |
| Data sharing | 50/month | Opt into usage data sharing |
$25 in free credits gets you roughly:
50/month data sharing is the most generous recurring credit in the API market. If you're not handling sensitive data, this effectively makes Grok free for small-to-medium workloads.
Full flagship comparison, per 1M tokens, April 2026:
| Model | Input | Output | Cache Hit | Context | Batch Output |
|---|---|---|---|---|---|
| Grok 4.1 Fast | $0.20 | $0.50 | $0.05 | 2M | $0.25 |
| Grok 4.20 | $2.00 | $6.00 | $0.20 | 2M | $3.00 |
| GPT-5.4 | $2.50 | 5.00 | $0.25 | 1.1M | $7.50 |
| GPT-5.4 Mini | $0.75 | $4.50 | $0.075 | 400K | $2.25 |
| Claude Sonnet 4.6 | $3.00 | 5.00 | $0.30 | 1M | $7.50 |
| Claude Opus 4.6 | $5.00 | $25.00 | $0.50 | 1M | 2.50 |
| DeepSeek V4 | $0.30 | $0.50 | $0.03 | 1M | N/A |
Key pricing insights from TokenMix.ai:
Grok 4.20 vs GPT-5.4: 20% cheaper input, 60% cheaper output. Grok wins decisively on cost for output-heavy tasks. Plus 2M context vs 1.1M.
Grok 4.1 Fast vs DeepSeek V4: Grok is 33% cheaper on input, identical on output, 2x the context. The budget tier winner depends on cache patterns — DeepSeek's $0.03 cache is cheaper than Grok's $0.05.
Grok 4.1 Fast vs GPT-5.4 Mini: Grok is 73% cheaper on input ($0.20 vs $0.75) and 89% cheaper on output ($0.50 vs $4.50). This is a massive gap at the mid-tier.
Context window dominance: Every Grok model has 2M tokens. No other provider offers this at any price point.
| Model | Monthly Cost (Cached) |
|---|---|
| Grok 4.1 Fast | $3.72 |
| DeepSeek V4 | $4.10 |
| GPT-5.4 Nano | $7.98 |
| Claude Haiku 4.5 | 5.60 |
Grok 4.1 Fast is the cheapest option — even beating DeepSeek V4 by 9%.
| Model | Standard | Cached |
|---|---|---|
| Grok 4.20 | $2,700 | $990 |
| GPT-5.4 | $5,625 | $2,063 |
| Claude Sonnet 4.6 | $5,850 | $2,138 |
| DeepSeek V4 | $285 | 40 |
Grok 4.20 saves ,073/month vs GPT-5.4 with caching — the output price advantage compounds at scale. DeepSeek remains cheapest overall.
| Model | Context | Can Handle 500K? | Input Cost/doc |
|---|---|---|---|
| Grok 4.20 | 2M | Yes | .00 |
| GPT-5.4 | 1.1M | Yes (with 2x surcharge) | $2.50* |
| Claude Opus 4.6 | 1M | Yes | $2.50 |
| DeepSeek V4 | 1M | Yes | $0.15 |
*GPT-5.4 hits 2x surcharge at 272K. Grok has no surcharge at any context length.
Price means nothing if quality is insufficient. Here's where Grok models stand:
| Benchmark | Grok 4.20 | Grok 4.1 Fast | GPT-5.4 | Claude Opus 4.6 | DeepSeek V4 |
|---|---|---|---|---|---|
| SWE-bench | ~78% | ~70% | 80% | 80.8% | 81% |
| MMLU | ~88% | ~82% | 90% | 89% | 88% |
| Coding (Arena) | Top 5 | Mid-tier | Top 3 | Top 2 | Top 5 |
Grok 4.20 is competitive but not leading. It sits 2-3 points behind GPT-5.4 and Claude Opus on SWE-bench. For most production tasks, this gap is negligible. For cutting-edge coding and reasoning, Claude Opus still leads.
Grok 4.1 Fast is noticeably weaker. ~70% SWE-bench puts it closer to GPT-4o level than GPT-5.4 level. Use it for simple tasks where the price advantage justifies the quality trade-off.
| Your Situation | Recommended Model | Why |
|---|---|---|
| Budget production, simple tasks | Grok 4.1 Fast ($0.20/$0.50) | Cheapest with 2M context |
| Output-heavy generation | Grok 4.20 ($2/$6) | 60% cheaper output than GPT/Claude |
| Need >1M context, no surcharge | Any Grok model | All have 2M flat pricing |
| Maximum quality, cost secondary | Claude Opus 4.6 | Still leads on benchmarks |
| Cost is everything | DeepSeek V4 | Cheapest overall, but only 1M context |
| Multi-model with auto-failover | Grok via TokenMix.ai | Unified API, route to cheapest available |
| Free prototyping | Grok ($25 free credits) | Most generous signup credits in the market |
Related: Compare all model pricing in our complete LLM API pricing comparison
Grok's pricing in 2026 is built around two advantages: the cheapest flagship output ($6/M vs 5/M from GPT and Claude) and the largest context window at every price tier (2M tokens across the board). For output-heavy workloads and long-context processing, Grok offers genuine savings of 40-60% compared to OpenAI and Anthropic.
The quality gap is real but narrowing. Grok 4.20 sits 2-3 points behind the leaders on SWE-bench, which is negligible for most production tasks. Grok 4.1 Fast is a tier below — use it for volume work, not quality-critical tasks.
The $25 signup credits and 50/month data sharing program make Grok effectively free for small teams. No other provider matches this generosity.
Compare Grok pricing against 155+ models in real time at tokenmix.ai/models.
Grok 4.1 Fast costs $0.20/M input and $0.50/M output — one of the cheapest APIs available. Grok 4.20 flagship costs $2.00/M input and $6.00/M output. All models include a 2M token context window at no surcharge.
Yes, significantly. Grok 4.20 output ($6/M) is 60% cheaper than GPT-5.4 ( 5/M). Grok 4.1 Fast ($0.20/$0.50) is 73-89% cheaper than GPT-5.4 Mini ($0.75/$4.50) across all dimensions.
Yes. New accounts receive $25 in free credits with no credit card required. An additional 50/month is available by opting into the data sharing program — the most generous recurring credit in the API market.
2 million tokens across every model — the largest in the industry. GPT-5.4 offers 1.1M, Claude and DeepSeek offer 1M. Grok charges flat pricing with no long-context surcharge.
Grok 4.1 Fast is 33% cheaper on input ($0.20 vs $0.30) and identical on output ($0.50). Grok has 2x the context (2M vs 1M). DeepSeek has cheaper cache hits ($0.03 vs $0.05) and slightly higher benchmark scores.
Grok 4.20 is competitive with GPT-5.4 and Claude Sonnet for most production tasks. It's 2-3 points behind on SWE-bench but offers better pricing and larger context. Grok 4.1 Fast is best suited for simpler tasks — not recommended for quality-critical coding.
Author: TokenMix Research Lab | Last Updated: April 2026 | Data Source: xAI Official Pricing, TokenMix.ai, and Artificial Analysis