Llama 4 Maverick API Rate Limits & Pricing (July 2026)

MetaGenerally available

API model string

meta-llama/Llama-4-Maverick-17B-128E-Instruct (host-dependent)

Context window

1M tokens (native); host-capped (Groq narrower; Fireworks ~1M)

Max output not published (host-capped)

Knowledge cutoff: not published
Released: Apr 2026
Modalities: text, image

Last verified July 10, 2026

Rate limits by tier

Llama 4 Maverick has no first-party hosted API from Meta (waitlist only). Rate limits are set entirely by the inference host. Groq offers a free tier at ~30 RPM; paid hosts (DeepInfra, Fireworks, Together AI) do not publicly publish their RPM limits — check your org's console.

Tier	Requirements	RPM	TPM	RPD	Concurrent	Notes
Meta first-party	Waitlist only — not an active commercial endpoint	not published	not published	not published	not published	Meta's own Llama API requires waitlist; Llama 4 Maverick is not an active first-party commercial endpoint as of July 2026
Groq (third-party)	Free tier available — no credit card required	~30 (free tier); higher on paid plans	not published	not published	not published	Context window capped below 1M native on Groq — check Groq console for exact limit. $0.50 in / $0.77 out per MTok (verified July 10, 2026).
DeepInfra (third-party)	Pay-as-you-go — credit balance required	not published	not published	not published	not published	$0.15 in / $0.60 out per MTok — cheapest available host (verified July 10, 2026). One of the most cost-effective options for Llama 4 Maverick.
Fireworks AI (third-party)	Pay-as-you-go	not published	not published	not published	not published	~$0.17 in / $0.60 out per MTok (verified July 10, 2026); ~1M context exposed — best option for long-context workloads near native 1M cap
Together AI (third-party)	Pay-as-you-go	not published	not published	not published	not published	~$0.55 in / $2.19 out per MTok (verify — significantly pricier than DeepInfra/Fireworks at ~3.6× input and ~3.7× output cost)

Swipe the table sideways to see every limit column.

1.All rate limits are set by the inference host, not Meta. Rate limits vary widely across hosts; always check the host's console for your org's current limits.
2.HTTP 429 returns with Retry-After header on all OpenAI-compatible hosts. Context windows are host-capped below the native 1M — Fireworks exposes ~1M; Groq's cap is narrower (check console).
3.Batch discounts are host-dependent: Amazon Bedrock offers ~50% off on eligible throughput; other hosts vary. Llama 4 Community License applies — platforms with >700M MAU require a separate Meta agreement.

Limits verified against the Meta docs, July 10, 2026.

Token pricing

What you pay per million tokens (USD). Input and output are billed separately.

Input

$0.15

per 1M tokens

Output

$0.60

per 1M tokens

Prices shown are for DeepInfra ($0.15/$0.60 per MTok) — the cheapest verified host as of July 10, 2026.
Groq: $0.50 in / $0.77 out (verified July 10, 2026). Fireworks: ~$0.17 in / $0.60 out (verified July 10, 2026). Together AI: ~$0.55 in / $2.19 out (verify — significantly pricier).
No standard cross-host prompt caching for Llama 4 Maverick. The 70%-cached scenario is not applicable. Cache system prompts at the application layer.
Amazon Bedrock offers batch inference at ~50% off for eligible throughput. Llama 4 Community License: platforms with >700M MAU need a separate Meta agreement.

Side-project (5M in / 1M out)

$1.35

per month

Assumptions

DeepInfra $0.15 in / $0.60 out — cheapest verified host; no caching standard

5M × $0.15 + 1M × $0.60 = $0.75 + $0.60 = $1.35. Groq alternative: 5M × $0.50 + 1M × $0.77 = $2.50 + $0.77 = $3.27.

Mid-scale app (60M in / 12M out)

$16.20

per month

Assumptions

DeepInfra $0.15/$0.60; no standard caching — 70%-cached scenario not applicable

60M × $0.15 + 12M × $0.60 = $9.00 + $7.20 = $16.20. No cache discount on Llama 4 hosts. Compare Llama 4 Scout on DeepInfra: $4.80 + $3.60 = $8.40 (48% cheaper).

High-volume (400M in / 80M out)

$108

per month

Assumptions

DeepInfra $0.15/$0.60; no standard caching

400M × $0.15 + 80M × $0.60 = $60 + $48 = $108. Compare Llama 4 Scout on DeepInfra: $32 + $24 = $56 — 48% cheaper. Use Maverick only if stronger reasoning justifies the premium.

Run your own numbers

Drag your real monthly token volumes and watch the bill update live — priced against rival models at the same usage.

Input tokens / month30M

1M500M

Output tokens / month5M

0.1M100M

Estimated meta-llama/Llama-4-Maverick-17B-128E-Instruct (host-dependent) spend

$7.50/mo

Input: $4.50

Output: $3.00

30M in × $0.150 + 5M out × $0.600 = $7.50

Same volume, priced across models

Llama 4 ScoutCheapest$3.90
meta-llama/Llama-4-Maverick-17B-128E-Instruct (host-dependent)This model$7.50
Kimi K2$49

Rivals priced at their published input/output rates for the same monthly volumes. Prompt caching is model-specific, so it is applied to meta-llama/Llama-4-Maverick-17B-128E-Instruct (host-dependent) only. Estimates for comparison; real bills vary with request shape and long-context surcharges.

meta-llama/Llama-4-Maverick-17B-128E-Instruct (host-dependent) vs the alternatives

Llama 4 Maverick compared to Llama 4 Scout (cheaper, lighter) and Kimi K2 (comparable MoE scale) as of July 2026.

Aspect	meta-llama/Llama-4-Maverick-17B-128E-Instruct (host-dependent)	Llama 4 Scout	Kimi K2
Input price (cheapest host)	$0.15/MTok (DeepInfra)	$0.08/MTok (DeepInfra)	$0.95/MTok (K2.6)
Output price (cheapest host)	$0.60/MTok (DeepInfra)	$0.30/MTok (DeepInfra)	$4.00/MTok (K2.6)
Context window	1M native (host-capped; ~1M on Fireworks)	10M native / 128K–1M hosted	262K (K2.6)
Reasoning quality	Stronger (128 experts, 400B total)	Lighter (16 experts, 109B total)	Strong (MoE ~1T)
Multimodal	Text + image	Text + image	Text + image
License	Llama 4 Community (>700M MAU needs Meta agreement)	Same	Modified MIT (permissive)
Cache pricing	null (host-dependent, not standard)	null (host-dependent)	$0.16/MTok cache-hit
Batch discount	~50% on Bedrock	~50% on Bedrock	~40%

Swipe the table sideways to see every model.

Hitting a 429? The playbook

The exact errors you'll see

429 Too Many Requestsrate_limit_exceededRateLimitErrorThrottlingException

Why it happens & how to fix it

Groq free tier RPM exhausted (~30 RPM limit on Llama 4 Maverick)

Upgrade to Groq paid tier, or shift large-context requests to DeepInfra ($0.15/MTok) or Fireworks (~$0.17/MTok) where RPM is typically higher for paying customers.

Host TPM cap hit during a traffic burst

Implement exponential backoff with jitter; split load across 2 hosts (e.g., DeepInfra + Fireworks) using round-robin routing with health-check fallback.

Context exceeds host's exposed window (some hosts cap below 1M native)

Move long-context requests to Fireworks which exposes ~1M context, or self-host Maverick for the full 1M native window.

Amazon Bedrock ThrottlingException — account quota limit reached

Go to AWS Console → Service Quotas → Amazon Bedrock → Llama 4 Maverick → request increase. Approval typically takes 1–3 business days.

Using Together AI at ~3.6× the DeepInfra price — budget exhausted at same token volume

Switch primary host to DeepInfra ($0.15/$0.60) or Fireworks (~$0.17/$0.60). Same model, same quality, significantly lower cost per token.

Retry strategy

Use exponential backoff with jitter. Honor the Retry-After header from the host — all OpenAI-compatible hosts return it on 429. Formula: wait = min(2^attempt + random(0,1), 60). Max 5 retries. With OpenAI SDK: set max_retries=5 to enable automatic retry handling.

retry.ts

1import OpenAI from 'openai';
2
3// Switch baseURL to your preferred host:
4// DeepInfra: https://api.deepinfra.com/v1/openai
5// Groq:      https://api.groq.com/openai/v1
6// Fireworks: https://api.fireworks.ai/inference/v1
7const client = new OpenAI({
8  baseURL: 'https://api.deepinfra.com/v1/openai',
9  apiKey: process.env.HOST_API_KEY,
10  maxRetries: 0, // handle retries manually
11});
12
13async function callWithRetry(
14  messages: OpenAI.Chat.ChatCompletionMessageParam[],
15  maxRetries = 5
16): Promise<string> {
17  for (let attempt = 0; attempt < maxRetries; attempt++) {
18    try {
19      const response = await client.chat.completions.create({
20        model: 'meta-llama/Llama-4-Maverick-17B-128E-Instruct',
21        messages,
22      });
23      return response.choices[0].message.content ?? '';
24    } catch (err: any) {
25      if (err?.status === 429) {
26        const retryAfter = parseInt(err?.headers?.['retry-after'] ?? '0', 10);
27        const jitter = Math.random();
28        const waitMs = retryAfter > 0
29          ? retryAfter * 1000
30          : Math.min(Math.pow(2, attempt) + jitter, 60) * 1000;
31        console.warn(`429 — attempt ${attempt + 1}/${maxRetries}, waiting ${Math.round(waitMs)}ms`);
32        await new Promise(r => setTimeout(r, waitMs));
33      } else {
34        throw err;
35      }
36    }
37  }
38  throw new Error('Max retries exceeded after repeated 429 responses');
39}

How to raise your limits

The ladder from the starter tier to enterprise — what each rung takes, and what it unlocks.

Free tier (Groq)

Immediate

Unlocks: ~30 RPM on Llama 4 Maverick; context capped below 1M; $0.50/$0.77 pricing

Pay-as-you-go (DeepInfra or Fireworks)

Immediate on payment

Add $10–$50 credit balance at deepinfra.com or fireworks.ai

Unlocks: Higher RPM, $0.15/$0.60 on DeepInfra, ~1M context on Fireworks; best cost/context combination

Together AI enterprise

3–5 business days

Contact Together AI sales for custom throughput agreement

Unlocks: Custom RPM/TPM, SLAs — though DeepInfra/Fireworks are significantly cheaper per token for the same model

Amazon Bedrock production

1–3 business days

Enable model access in Bedrock console → request throughput increase via AWS Service Quotas → Amazon Bedrock → Llama 4 Maverick

Unlocks: Batch inference ~50% off, AWS ecosystem integration, SLAs, enterprise compliance

Self-host

Days to weeks

Download open weights under Llama 4 Community License; deploy on 8× H100 or A100 (17B active parameters — MoE architecture is memory-efficient vs dense 405B)

Unlocks: Full 1M context, zero per-token cost, complete control. Requires Meta agreement if platform has >700M MAU.

Cut your token spend

Choose DeepInfra or Fireworks over Together AI

~3.6× cheaper input ($0.15 vs $0.55) and ~3.7× cheaper output ($0.60 vs $2.19)

Update your base_url to https://api.deepinfra.com/v1/openai and use DeepInfra's Maverick model string. Same model quality at a fraction of the cost.

Route lighter tasks to Llama 4 Scout

~47% cheaper on input ($0.08 vs $0.15) and 50% cheaper on output ($0.30 vs $0.60)

Use Scout for classification, summarization, and extraction tasks. Reserve Maverick for tasks needing stronger multi-step reasoning or complex instruction following.

Use Amazon Bedrock batch for bulk inference

~50% off on eligible throughput

Submit document processing, evals, and bulk classification jobs as Bedrock batch requests rather than synchronous API calls. Ideal for non-real-time workloads.

Keep requests within the host's context cap

Avoids context errors and maximizes throughput on shared infrastructure

Groq caps Maverick below 1M native. Use Fireworks (~1M exposed) for genuinely long-context workloads. Chunk documents to fit within the actual host limit.

Use Fireworks for 1M context, DeepInfra for cost

Match the right host to the workload and avoid overpaying or hitting context limits

Requests under 320K tokens: use DeepInfra at $0.15/MTok. Requests requiring up to 1M tokens: use Fireworks at ~$0.17/MTok. Only pay the Fireworks premium when you need the extra context.

Cache system prompts at the application layer

Typically 20–40% reduction in repeated token spend

No native cross-host caching standard exists for Llama 4 Maverick. Store and reuse system prompt tokens in your backend so they are not re-sent with every request.

Frequently asked questions

What is Llama 4 Maverick's API rate limit?

Llama 4 Maverick has no first-party Meta API. Rate limits are set by your inference host. Groq's free tier is approximately 30 RPM. DeepInfra, Fireworks, and Together AI do not publish specific RPM limits — check your organization's console on each platform. All hosts return 429 with a Retry-After header when limits are exceeded.

Is Llama 4 Maverick free to use?

Groq offers Llama 4 Maverick on a free tier (~30 RPM, context capped) at $0.50/$0.77 per MTok. There is no zero-cost free inference tier beyond Groq's free plan. For cost-sensitive workloads, DeepInfra at $0.15/$0.60 per MTok is the cheapest paid option (verified July 10, 2026).

Llama 4 Maverick vs Llama 4 Scout: which should I use?

Llama 4 Scout is ~47% cheaper on input and 50% cheaper on output, and it has a 10M native context window. Use Scout for classification, summarization, extraction, and most conversational tasks. Use Maverick when your workload genuinely requires stronger multi-step reasoning (128 experts, 400B total parameters). Route-and-split: use Scout by default, escalate to Maverick for complex tasks.

How do I get the cheapest price for Llama 4 Maverick?

DeepInfra offers the lowest verified price: $0.15 in / $0.60 out per MTok as of July 10, 2026. That is 3.6× cheaper on input than Together AI ($0.55/MTok). Fireworks is close at ~$0.17/$0.60 and offers ~1M context — useful if you need more than DeepInfra's exposed window.

Does Llama 4 Maverick support the full 1M context window?

Llama 4 Maverick has a 1M native context window, but hosted access is capped by the inference provider. Fireworks exposes approximately 1M context — the closest to native of any listed host. Groq caps the context window below 1M (check the Groq console for the current limit). For workloads requiring the full 1M window, Fireworks or self-hosting are your options.

How do I increase Llama 4 Maverick rate limits?

Rate limits are controlled by your inference host. To increase them: upgrade from Groq's free tier to a paid plan; add credit balance on DeepInfra or Fireworks for higher throughput; request a quota increase on Amazon Bedrock via AWS Service Quotas (1–3 business days); or contact Together AI sales for enterprise throughput. If you need help designing a multi-host routing architecture for higher effective limits, the RapidDev team offers free scoping calls at rapidevelopers.com/contact.

What is the Llama 4 Community License?

Llama 4 Maverick is released under the Llama 4 Community License, which allows free commercial use and self-hosting. However, platforms with more than 700 million monthly active users must obtain a separate license agreement directly from Meta. For most applications, this restriction does not apply — but it is an important consideration for large-scale consumer products.

We build AI apps that don't hit rate limits

Retry, backoff & caching built in
Multi-provider fallback routing
Fixed price, you own the code

Get a free estimate

30-min call. No commitment.

Llama 4 Maverick API Rate Limits, Pricing & Performance (July 2026)

Rate limits by tier

Token pricing

Run your own numbers

meta-llama/Llama-4-Maverick-17B-128E-Instruct (host-dependent) vs the alternatives

Hitting a 429? The playbook

How to raise your limits

Free tier (Groq)

Pay-as-you-go (DeepInfra or Fireworks)

Together AI enterprise

Amazon Bedrock production

Self-host

Cut your token spend

Choose DeepInfra or Fireworks over Together AI

Route lighter tasks to Llama 4 Scout

Use Amazon Bedrock batch for bulk inference

Keep requests within the host's context cap

Use Fireworks for 1M context, DeepInfra for cost

Cache system prompts at the application layer

Frequently asked questions

We build AI apps that don't hit rate limits

Your next step

Still weighing your options?

We put the rapid in RapidDev

Llama 4 Maverick API Rate Limits, Pricing & Performance (July 2026)

Rate limits by tier

Token pricing

Run your own numbers

meta-llama/Llama-4-Maverick-17B-128E-Instruct (host-dependent) vs the alternatives

Hitting a 429? The playbook

How to raise your limits

Free tier (Groq)

Pay-as-you-go (DeepInfra or Fireworks)

Together AI enterprise

Amazon Bedrock production

Self-host

Cut your token spend

Choose DeepInfra or Fireworks over Together AI

Route lighter tasks to Llama 4 Scout

Use Amazon Bedrock batch for bulk inference

Keep requests within the host's context cap

Use Fireworks for 1M context, DeepInfra for cost

Cache system prompts at the application layer

Related models

Llama 4 Scout API Rate Limits & Pricing

Llama 3.1 API Rate Limits & Pricing

Llama 3 API Rate Limits & Pricing

More meta-llama/Llama-4-Maverick-17B-128E-Instruct (host-dependent) resources

Llama 4 Maverick via OpenAI-Compatible Wrapper in n8n & Make

Llama 4 Maverick 429 Errors on Groq & ThrottlingException on Bedrock

Frequently asked questions

We build AI apps that don't hit rate limits

Your next step

Still weighing your options?

We put the rapid in RapidDev