Llama 4 Scout (Meta) vs Llama 3.3 70B (Together)
Side-by-side comparison of API pricing, specs, benchmarks, and capabilities
| Specification |
by Meta
3 providers:
Groq $0.110/$0.340
Meta $0.110/$0.340
DeepInfra $0.100/$0.300
|
by Together
2 providers:
Meta $0.590/$0.790
Together $1.04/$1.04
|
|---|---|---|
| Overview | ||
| Status | Current open weights Open weights | Current open weights Open weights |
| Released | Apr 6, 2025 | Dec 6, 2024 |
| Pricing per million tokens | ||
| Input | $0.110/Mtok | $1.04/Mtok |
| Output | $0.340/Mtok | $1.04/Mtok |
| Blended avg | $0.168/Mtok | $1.04/Mtok |
| Specifications | ||
| Context window | 10M tokens | 128K tokens |
| Parameters | 17B (16 experts) | 70B |
| Speed (TPS) | — | — |
| Modalities | ||
| Input |
textimage
|
text
|
| Benchmarks sources: Vellum LLM Leaderboard / Vellum LLM Leaderboard (Jun 2026), Kaggle dataset | ||
| Avg benchmark score | 46.8 | 46.6 |
| Perf / dollar | 279.4 | 44.8 |
| GPQA Diamond | 57.2 | 50.5 |
| SWE-Bench Verified | 28 | 22 |
| Humanity's Last Exam | 12 | 10 |
| ARC-AGI 2 | 15 | 12 |
| AIME 2025 | 45 | 38 |
| MMMLU | 75 | 72 |
| BFCL | 58 | 77.3 |
| HumanEval | 80 | 78 |
| MATH 500 | 65 | 60 |
| LiveCodeBench | 32.8 | — |
| Independent composite scores each on its own scale — not part of the average above | ||
| AA Intelligence Index | 10.3 | — |
| AA Agentic Index | 1.1 | — |
| AA-Omniscience Index (−100–100) | -52.1 | — |
| Providers | ||
| Available from |
Groq — $0.110/$0.340/Mtok
Meta — $0.110/$0.340/Mtok
DeepInfra — $0.100/$0.300/Mtok
|
Meta — $0.590/$0.790/Mtok
Together — $1.04/$1.04/Mtok
|
Cost at scale
1M tokens · 50/50 input/output| Volume | Llama 4 Scout | Llama 3.3 70B | Savings |
|---|---|---|---|
| 1M tokens | $0.17 | $1.04 | $0.87 (83.7%) |
| 10M tokens | $1.68 | $10.4 | $8.73 (83.9%) |
| 100M tokens | $16.75 | $104 | $87.25 (83.9%) |
| 1000M tokens | $167.5 | $1040 | $872.5 (83.9%) |
When to pick which
distilled from the pricing and spec data above- cost dominates: $0.168/Mtok blended vs $1.04 — 83.9% less on the same 50/50 token mix
- you need the longer context: 10M tokens vs 128K (78.1×)
- raw capability matters more than price: 46.8 vs 46.6 average benchmark score
- provider flexibility: available from 3 providers, so you can price-shop hosts or fail over
- this pairing gives Llama 3.3 70B (Together) no edge on price, context, cache discount, or measured performance — pick it only for qualitative fit (ecosystem, compliance, model behavior on your prompts)
Price movement
recorded changes since we began tracking each model · last checked Aug 9, 2026| Model | Changes | Latest move | Old → new $/M |
|---|---|---|---|
| Llama 4 Scout | 0 | — no move | Stable since we began tracking — no change recorded |
| Llama 3.3 70B | 1 | price rise on Jun 25, 2026 | $0.88/$0.88 → $1.04/$1.04/Mtok |
Summary
Llama 4 Scout by Meta costs $0.110/Mtok input and $0.340/Mtok output, with a 10M-token context window. It supports text, image input and is available from 3 providers.
Llama 3.3 70B by Together costs $1.04/Mtok input and $1.04/Mtok output, with a 128K-token context window. It supports text input and is available from 2 providers.
On a blended cost basis, Llama 4 Scout is 83.9% cheaper than Llama 3.3 70B.It also has a larger context window.
On benchmarks, Llama 4 Scout scores higher (46.8 vs 46.6) on average. In terms of value, Llama 4 Scout has better performance per dollar (279.4 vs 44.8).
Note: Pricing is per million tokens. Actual costs vary with usage patterns, prompt caching, and batch discounts. Always verify against official provider pricing pages.