Skip to content
CostPerPrompt

Llama 3.3 70B Instruct API Pricing

Meta · context window 131K · prices updated 2026-11-22

Input per 1M tokens
$0.71
Output per 1M tokens
$0.71
Cached input per 1M
$0.71
0% cheaper than fresh input

Where Llama 3.3 70B Instruct sits on price

At $0.71 per million output tokens, Llama 3.3 70B Instruct is a mid tier model — 3.5× cheaper than the median output price of $2.50, which puts it cheaper than 76% and pricier than 24% of the models we track. It sits 5th cheapest of the 7 Meta models we track. Output is only 1.0× the input price, an unusually flat spread: long prompts hurt about as much as long answers, so prompt size is where the savings are.

Output tokens per $1
1,408,451
One full context fill
$0.0931
Cheaper than
76% of tracked models

What real workloads cost on Llama 3.3 70B Instruct

These three workloads are the ones teams actually run on a mid-priced model — a $0.71/1M model is not bought for the same job as one ten times the price.

Workload Per request Per month
Support chatbot — 500 conversations/day (5K in / 1.4K out) Conversation history re-sent each turn — the classic prompt-caching win. $0.0045 $68.16
Document pipeline — 10K docs/day (3K in / 500 out) Batch-friendly: no user waiting, so a ~50% batch discount usually applies. $0.0025 $745
Small app — 1K requests/day (500 in / 150 out) The baseline most side projects actually run at. $0.0005 $13.85

Model your exact traffic in the API cost calculator — it preloads Llama 3.3 70B Instruct with caching and batch options.

Llama 3.3 70B Instruct price history

Meta last cut Llama 3.3 70B Instruct's output price on Nov 10, 2026 (-10% per output token), from $0.59/$0.79 to $0.71/$0.71 per 1M in/out. That is one of 15 repricings in the 112 days we have tracked it — worth knowing before you hard-code today's rate into a budget.

Date Input /1M Output /1M Output change
Aug 2, 2026 (tracking began) $0.13 $0.4 —
Aug 4, 2026 $0.1 $0.32 -20%
Aug 26, 2026 $0.71 $0.71 +122%
Sep 3, 2026 $0.1 $0.32 -55%
Oct 4, 2026 $0.22 $0.5 +56%
Oct 5, 2026 $0.1 $0.32 -36%
Oct 6, 2026 $0.22 $0.5 +56%
Oct 8, 2026 $0.1 $0.32 -36%
Oct 9, 2026 $0.22 $0.5 +56%
Oct 13, 2026 $0.59 $0.79 +58%
Oct 16, 2026 $0.1 $0.32 -59%
Nov 3, 2026 $0.2 $0.52 +63%
Nov 4, 2026 $0.1 $0.32 -38%
Nov 5, 2026 $0.2 $0.52 +63%
Nov 6, 2026 $0.59 $0.79 +52%
Nov 10, 2026 $0.71 $0.71 -10%

Change-points from our daily price snapshots (tracking since Aug 2, 2026; intraday moves between snapshots are not captured).

Llama 3.3 70B Instruct vs Qwen3 VL 8B Instruct

The closest-priced alternative from another vendor is Qwen3 VL 8B Instruct (Alibaba (Qwen)) — 6% more expensive on output, with $0.46 less per million input tokens. When two models land this close on price, the decision is quality on your own workload, not the price sheet: run 50 real requests through both and compare.

See Qwen3 VL 8B Instruct pricing →

Cheaper alternatives

More Meta models

Frequently asked questions

› How much does the Llama 3.3 70B Instruct API cost?

Llama 3.3 70B Instruct costs $0.71 per million input tokens and $0.71 per million output tokens, with cached input at $0.71 per million (0% cheaper). That works out to roughly 1,408,451 output tokens per dollar.

› What does the support chatbot workload cost on Llama 3.3 70B Instruct?

Support chatbot — 500 conversations/day (5K in / 1.4K out) costs about $0.0045 per request and $68.16 per month on Llama 3.3 70B Instruct. Conversation history re-sent each turn — the classic prompt-caching win.

› What does it cost to fill Llama 3.3 70B Instruct's 131K context window?

Sending 131K of input in a single request costs $0.0931 at $0.71 per million tokens — before any output. There is no cached-input tier on this model, so every repeat of that context is billed at full price. This is why large context windows are cheap to advertise and expensive to actually use.

› Is Llama 3.3 70B Instruct worth the price?

Llama 3.3 70B Instruct sits in the middle of the market (5 of 7 by price within Meta). The honest test is a routing experiment: send the same 200 real requests to this model and to a tier below, and compare failure rate against the price gap — most teams find a majority of traffic never needed the pricier model.