// Meta
Llama 3.3 70B Instruct
A Meta model. 5 prices across 5 channels side by side, each with its source and check date – plus specs, licence, speed and benchmarks.
- Official in/out per 1M
- —
- Lowest (Novita)
- $0.135 / $0.4
- Context
- 12K
- AA index
- 7.7
Prices checked 2026-10-22Official sources linked on every rowNo estimatesCheapest LLM APIs →
01Specs at a glance
Official values first; gaps filled from OpenRouter’s API where the vendor page is silent.
| Vendor | Meta |
|---|---|
| Context window | 12K (12,288 tokens) |
| Max output | 11K |
| Input | Text |
| Output | Text |
| Tool calling / JSON | Yes / Yes |
| Weights & licence | Open weights · Llama 3.3 Community · Hugging Face gated – request access |
| Parameters | 70.6B (from weight files) |
| Released | Dec 6, 2024 |
| Model IDs | meta-llama/llama-3.3-70b-instruct |
02Prices on 5 channels
Per 1M tokens, in the currency each source publishes (1 USD = 6.7 CNY for the blended column). Sorted: official first, then cheapest.
| Channel | Source | Input | Output | Cache hit | Blended $ | Context | Notes |
|---|---|---|---|---|---|---|---|
| Novita · bf16 Via OpenRouter / provider listed there | OpenRouter 10-22 | $0.135 | $0.4 | — | 0.201 | 12K | Novita price as listed on OpenRouter (pass-through; quantization=bf16; discount field=0); not checked against Novita's own page |
| Parasail · fp8 Via OpenRouter / provider listed there | OpenRouter 10-22 | $0.22 | $0.5 | $0.11 | 0.29 | 131K | Parasail price as listed on OpenRouter (pass-through; quantization=fp8; discount field=0); not checked against Parasail's own page |
| SambaNova · unknown Via OpenRouter / provider listed there | OpenRouter 10-22 | $0.45 | $0.9 | — | 0.563 | 131K | SambaNova price as listed on OpenRouter (pass-through; quantization=unknown; discount field=0.25); not checked against SambaNova's own page |
| Groq · unknown Via OpenRouter / provider listed there | OpenRouter 10-22 | $0.59 | $0.79 | $0.295 | 0.64 | 131K | Groq price as listed on OpenRouter (pass-through; quantization=unknown; discount field=0); not checked against Groq's own page |
| Together · unknown Via OpenRouter / provider listed there | OpenRouter 10-22 | $1.04 | $1.04 | — | 1.04 | 131K | Together price as listed on OpenRouter (pass-through; quantization=unknown; discount field=0); not checked against Together's own page |
“Official” = vendor pricing page; “Platform” = the platform’s own pricing page; “OpenRouter” = price listed on OpenRouter, not checked on the platform’s site. Click a source tag to open the original page.
03Where you can use it
Derived only from the price sources above. Country restrictions are set by each vendor – check their terms before you sign up.
- Via OpenRouter / provider listed there
Novita · Parasail · SambaNova · Groq · Together - Self-host the open weights
GPU / VRAM sizing and build plans→ - Free option
Llama-3.3-70B-Instruct→
04Performance
All results →Speed and AA index from Artificial Analysis public leaderboards (not our own tests); benchmark scores from our benchmark library, best result per benchmark.
- AA indexLlama 3.3 Instruct 70B
- 7.7
- Output speed (median)
- 93 tok/s
- Time to first token
- 1.77s
- AA index per $ (blended)
- 38.08
| Benchmark | Best score | Source | Date |
|---|---|---|---|
| LMArena Text | 1318 | official | 2026-10-08 |
| Open LLM Leaderboard v2 | 44.9% | official | 2025-03-13 |
| GPQA Diamond | 47.4% | independent | 2025-01-27 |
| MMLU | 86.3% | paper | 2024-12-06 |
| MMLU-Pro | 65.9% | official | 2026-03-11 |
| AIME (OTIS Mock AIME 2024–2025) | 5.1% | independent | 2025-02-25 |
| MATH Level 5 | 41.6% | independent | 2025-01-27 |
| BFCL v4 (function calling) | 31.9% | official | 2026-10-17 |
05Price changes we recorded
All price changes →From our daily re-checks of the sources above.
No price change recorded since we started tracking (2026-10-10).
06More from Meta
FAQ
How much does the Llama 3.3 70B Instruct API cost?
There is no official pay-as-you-go price in our data. The lowest listed price is $0.135 / $0.4 on Novita.
What is the context window of Llama 3.3 70B Instruct?
12K tokens of context, up to 11K output tokens. Platforms may cap it lower – see the table.
Is Llama 3.3 70B Instruct open source?
Yes, the weights are published on Hugging Face (meta-llama/Llama-3.3-70B-Instruct). Licence: Llama 3.3 Community.