Llama 3.3 70B
Open-weightLast verified: June 2026· Meta (via hosts) Llama 3.3 70B pricing
Open-weight model; hosted prices vary widely by provider. The figures shown are a representative hosted rate.
- Input / 1M
- $0.20
- Output / 1M
- $0.60
- Context
- 128K tokens
- Provider
- Meta (via hosts)
Pricing verified June 2026. Prices change frequently. Always confirm against the provider’s official pricing page before relying on these figures for budgeting. Official pricing →
When to choose Llama 3.3 70B
The row where the per-token number matters least. Llama 3.3 70B is open-weight, so the real decision is self-host versus a hosted API rather than $0.20 against somebody else’s $0.27. The figures shown are a representative hosted rate and vary widely by provider; self-hosting swaps the per-token line for fixed infrastructure that only pays off above a volume worth working out before you commit. Choose it when data residency, air-gapping, or high steady-state volume are driving the decision, and a hosted API when they are not.
What Llama 3.3 70B costs in practice
At the representative hosted rate a 50M/10M month is $16, and filling the 128K window costs under three cents of input. Treat both as indicative rather than a quote: hosted Llama pricing varies widely by provider, and the self-hosted alternative replaces per-token billing entirely with GPU hours that do not scale with tokens at all. The crossover between those two shapes is a real calculation with your own volume in it, and it is the one worth doing before anyone argues about a few cents per million tokens.
| Scenario | Input / mo | Output / mo | Est. cost / mo |
|---|---|---|---|
| Prototype | 2M | 0.5M | $1 |
| Growing product | 50M | 10M | $16 |
| At scale | 500M | 100M | $160 |
Plug in your own numbers with the cost calculator.
Where Llama 3.3 70B fits in a routing setup
The routing decision for this row is unusual, because one of the options is not a row at all. Hosted, it does the reference month for $16, close enough to DeepSeek-V3’s $24.50 that either could take the same traffic. Self-hosted, the per-token line disappears and the comparison becomes GPU hours against token spend, which only favours self-hosting above a volume you can work out from your own traces. That calculation belongs before the architecture decision rather than after it.
Cheaper alternatives to Llama 3.3 70B
The models immediately below Llama 3.3 70B on blended cost, nearest first. A swap only pays if quality holds on the routes you move — see the LLM cost optimization playbook for how to prove that before you move them.
Frequently asked questions
How much does Llama 3.3 70B cost?
Llama 3.3 70B bills $0.20 per million input tokens and $0.60 per million output tokens — $0.30 per million blended at a three-to-one input:output split. A workload of 50M input and 10M output tokens per month works out at about $16. Confirm current pricing with the provider.
What is Llama 3.3 70B's context window?
Llama 3.3 70B has a 128K-token context window. One request that fills the whole window costs about $0.03 at the input rate, before the reply is billed.
Where does Llama 3.3 70B sit on price?
At $0.30 per million blended — three input tokens for every output token — Llama 3.3 70B is dearer than 1 of the 19 models tracked here and cheaper than 17. Output bills at 3.0x the input rate, and the nearest cheaper model on this table is GPT-4o mini at $0.26 blended.
Related concepts: context window, model routing, prompt caching, evals.