Llama 3.3 70B Instruct
Meta · Released Dec 2024
Meta's 70-billion-parameter instruction-tuned language model with a 131k token context window, released after Llama 3.1 70B Instruct.
- Strengths
- Handles long-context tasks and maintains performance on complex reasoning, coding, and multi-turn dialogue at its 70B scale.
- Best for
- Applications requiring sustained context over long documents, extended conversations, or complex problem-solving where a larger model capacity is needed.
- Limitations
- Larger parameter count means higher latency and throughput requirements compared to smaller models in Meta's lineup; later models like Llama 4 Maverick offer expanded capability if agentic reasoning or tool use is required.
Input / 1M
$0.13
Output / 1M
$0.4
Cached input / 1M
—
Context window
131K
Price history
Input (solid)Output (dashed)
Price change
- 30d
- in increased 30.0% out increased 25.0%
- 90d
- in increased 30.0% out increased 25.0%
- 1y
- in increased 30.0% out increased 25.0%
- Since launch
- in increased 30.0% out increased 25.0%
Snapshots
| Effective | Input | Output | Cached in | Note | Source |
|---|---|---|---|---|---|
| 15 Jul 2026 | $0.13 | $0.4 | — | Imported from OpenRouter | openrouter.ai |
| 11 Jun 2026 | $0.1 | $0.32 | — | Imported from OpenRouter | openrouter.ai |